Showing preview only (302K chars total). Download the full file or copy to clipboard to get everything.
Repository: elyha7/yoloface
Branch: master
Commit: 04bfe3459db5
Files: 43
Total size: 288.1 KB
Directory structure:
gitextract_uz5iw4_8/
├── LICENSE
├── README.md
├── face_detector.py
├── models/
│ ├── __init__.py
│ ├── common.py
│ ├── experimental.py
│ ├── export.py
│ ├── yolo.py
│ ├── yolov5-0.5.yaml
│ ├── yolov5l.yaml
│ ├── yolov5l6.yaml
│ ├── yolov5m.yaml
│ ├── yolov5m6.yaml
│ ├── yolov5n.yaml
│ ├── yolov5n6.yaml
│ ├── yolov5s.yaml
│ └── yolov5s6.yaml
├── requirements.txt
├── utils/
│ ├── __init__.py
│ ├── activations.py
│ ├── autoanchor.py
│ ├── aws/
│ │ ├── __init__.py
│ │ ├── mime.sh
│ │ ├── resume.py
│ │ └── userdata.sh
│ ├── datasets.py
│ ├── face_datasets.py
│ ├── general.py
│ ├── google_app_engine/
│ │ ├── Dockerfile
│ │ ├── additional_requirements.txt
│ │ └── app.yaml
│ ├── google_utils.py
│ ├── infer_utils.py
│ ├── loss.py
│ ├── metrics.py
│ ├── plots.py
│ ├── preprocess_utils.py
│ ├── torch_utils.py
│ └── wandb_logging/
│ ├── __init__.py
│ ├── log_dataset.py
│ └── wandb_utils.py
└── weights/
├── download_weights.sh
└── yolov5n_state_dict.pt
================================================
FILE CONTENTS
================================================
================================================
FILE: LICENSE
================================================
GNU GENERAL PUBLIC LICENSE
Version 3, 29 June 2007
Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/>
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The GNU General Public License is a free, copyleft license for
software and other kinds of works.
The licenses for most software and other practical works are designed
to take away your freedom to share and change the works. By contrast,
the GNU General Public License is intended to guarantee your freedom to
share and change all versions of a program--to make sure it remains free
software for all its users. We, the Free Software Foundation, use the
GNU General Public License for most of our software; it applies also to
any other work released this way by its authors. You can apply it to
your programs, too.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
them if you wish), that you receive source code or can get it if you
want it, that you can change the software or use pieces of it in new
free programs, and that you know you can do these things.
To protect your rights, we need to prevent others from denying you
these rights or asking you to surrender the rights. Therefore, you have
certain responsibilities if you distribute copies of the software, or if
you modify it: responsibilities to respect the freedom of others.
For example, if you distribute copies of such a program, whether
gratis or for a fee, you must pass on to the recipients the same
freedoms that you received. You must make sure that they, too, receive
or can get the source code. And you must show them these terms so they
know their rights.
Developers that use the GNU GPL protect your rights with two steps:
(1) assert copyright on the software, and (2) offer you this License
giving you legal permission to copy, distribute and/or modify it.
For the developers' and authors' protection, the GPL clearly explains
that there is no warranty for this free software. For both users' and
authors' sake, the GPL requires that modified versions be marked as
changed, so that their problems will not be attributed erroneously to
authors of previous versions.
Some devices are designed to deny users access to install or run
modified versions of the software inside them, although the manufacturer
can do so. This is fundamentally incompatible with the aim of
protecting users' freedom to change the software. The systematic
pattern of such abuse occurs in the area of products for individuals to
use, which is precisely where it is most unacceptable. Therefore, we
have designed this version of the GPL to prohibit the practice for those
products. If such problems arise substantially in other domains, we
stand ready to extend this provision to those domains in future versions
of the GPL, as needed to protect the freedom of users.
Finally, every program is threatened constantly by software patents.
States should not allow patents to restrict development and use of
software on general-purpose computers, but in those that do, we wish to
avoid the special danger that patents applied to a free program could
make it effectively proprietary. To prevent this, the GPL assures that
patents cannot be used to render the program non-free.
The precise terms and conditions for copying, distribution and
modification follow.
TERMS AND CONDITIONS
0. Definitions.
"This License" refers to version 3 of the GNU General Public License.
"Copyright" also means copyright-like laws that apply to other kinds of
works, such as semiconductor masks.
"The Program" refers to any copyrightable work licensed under this
License. Each licensee is addressed as "you". "Licensees" and
"recipients" may be individuals or organizations.
To "modify" a work means to copy from or adapt all or part of the work
in a fashion requiring copyright permission, other than the making of an
exact copy. The resulting work is called a "modified version" of the
earlier work or a work "based on" the earlier work.
A "covered work" means either the unmodified Program or a work based
on the Program.
To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy. Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.
To "convey" a work means any kind of propagation that enables other
parties to make or receive copies. Mere interaction with a user through
a computer network, with no transfer of a copy, is not conveying.
An interactive user interface displays "Appropriate Legal Notices"
to the extent that it includes a convenient and prominently visible
feature that (1) displays an appropriate copyright notice, and (2)
tells the user that there is no warranty for the work (except to the
extent that warranties are provided), that licensees may convey the
work under this License, and how to view a copy of this License. If
the interface presents a list of user commands or options, such as a
menu, a prominent item in the list meets this criterion.
1. Source Code.
The "source code" for a work means the preferred form of the work
for making modifications to it. "Object code" means any non-source
form of a work.
A "Standard Interface" means an interface that either is an official
standard defined by a recognized standards body, or, in the case of
interfaces specified for a particular programming language, one that
is widely used among developers working in that language.
The "System Libraries" of an executable work include anything, other
than the work as a whole, that (a) is included in the normal form of
packaging a Major Component, but which is not part of that Major
Component, and (b) serves only to enable use of the work with that
Major Component, or to implement a Standard Interface for which an
implementation is available to the public in source code form. A
"Major Component", in this context, means a major essential component
(kernel, window system, and so on) of the specific operating system
(if any) on which the executable work runs, or a compiler used to
produce the work, or an object code interpreter used to run it.
The "Corresponding Source" for a work in object code form means all
the source code needed to generate, install, and (for an executable
work) run the object code and to modify the work, including scripts to
control those activities. However, it does not include the work's
System Libraries, or general-purpose tools or generally available free
programs which are used unmodified in performing those activities but
which are not part of the work. For example, Corresponding Source
includes interface definition files associated with source files for
the work, and the source code for shared libraries and dynamically
linked subprograms that the work is specifically designed to require,
such as by intimate data communication or control flow between those
subprograms and other parts of the work.
The Corresponding Source need not include anything that users
can regenerate automatically from other parts of the Corresponding
Source.
The Corresponding Source for a work in source code form is that
same work.
2. Basic Permissions.
All rights granted under this License are granted for the term of
copyright on the Program, and are irrevocable provided the stated
conditions are met. This License explicitly affirms your unlimited
permission to run the unmodified Program. The output from running a
covered work is covered by this License only if the output, given its
content, constitutes a covered work. This License acknowledges your
rights of fair use or other equivalent, as provided by copyright law.
You may make, run and propagate covered works that you do not
convey, without conditions so long as your license otherwise remains
in force. You may convey covered works to others for the sole purpose
of having them make modifications exclusively for you, or provide you
with facilities for running those works, provided that you comply with
the terms of this License in conveying all material for which you do
not control copyright. Those thus making or running the covered works
for you must do so exclusively on your behalf, under your direction
and control, on terms that prohibit them from making any copies of
your copyrighted material outside their relationship with you.
Conveying under any other circumstances is permitted solely under
the conditions stated below. Sublicensing is not allowed; section 10
makes it unnecessary.
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
No covered work shall be deemed part of an effective technological
measure under any applicable law fulfilling obligations under article
11 of the WIPO copyright treaty adopted on 20 December 1996, or
similar laws prohibiting or restricting circumvention of such
measures.
When you convey a covered work, you waive any legal power to forbid
circumvention of technological measures to the extent such circumvention
is effected by exercising rights under this License with respect to
the covered work, and you disclaim any intention to limit operation or
modification of the work as a means of enforcing, against the work's
users, your or third parties' legal rights to forbid circumvention of
technological measures.
4. Conveying Verbatim Copies.
You may convey verbatim copies of the Program's source code as you
receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy an appropriate copyright notice;
keep intact all notices stating that this License and any
non-permissive terms added in accord with section 7 apply to the code;
keep intact all notices of the absence of any warranty; and give all
recipients a copy of this License along with the Program.
You may charge any price or no price for each copy that you convey,
and you may offer support or warranty protection for a fee.
5. Conveying Modified Source Versions.
You may convey a work based on the Program, or the modifications to
produce it from the Program, in the form of source code under the
terms of section 4, provided that you also meet all of these conditions:
a) The work must carry prominent notices stating that you modified
it, and giving a relevant date.
b) The work must carry prominent notices stating that it is
released under this License and any conditions added under section
7. This requirement modifies the requirement in section 4 to
"keep intact all notices".
c) You must license the entire work, as a whole, under this
License to anyone who comes into possession of a copy. This
License will therefore apply, along with any applicable section 7
additional terms, to the whole of the work, and all its parts,
regardless of how they are packaged. This License gives no
permission to license the work in any other way, but it does not
invalidate such permission if you have separately received it.
d) If the work has interactive user interfaces, each must display
Appropriate Legal Notices; however, if the Program has interactive
interfaces that do not display Appropriate Legal Notices, your
work need not make them do so.
A compilation of a covered work with other separate and independent
works, which are not by their nature extensions of the covered work,
and which are not combined with it such as to form a larger program,
in or on a volume of a storage or distribution medium, is called an
"aggregate" if the compilation and its resulting copyright are not
used to limit the access or legal rights of the compilation's users
beyond what the individual works permit. Inclusion of a covered work
in an aggregate does not cause this License to apply to the other
parts of the aggregate.
6. Conveying Non-Source Forms.
You may convey a covered work in object code form under the terms
of sections 4 and 5, provided that you also convey the
machine-readable Corresponding Source under the terms of this License,
in one of these ways:
a) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by the
Corresponding Source fixed on a durable physical medium
customarily used for software interchange.
b) Convey the object code in, or embodied in, a physical product
(including a physical distribution medium), accompanied by a
written offer, valid for at least three years and valid for as
long as you offer spare parts or customer support for that product
model, to give anyone who possesses the object code either (1) a
copy of the Corresponding Source for all the software in the
product that is covered by this License, on a durable physical
medium customarily used for software interchange, for a price no
more than your reasonable cost of physically performing this
conveying of source, or (2) access to copy the
Corresponding Source from a network server at no charge.
c) Convey individual copies of the object code with a copy of the
written offer to provide the Corresponding Source. This
alternative is allowed only occasionally and noncommercially, and
only if you received the object code with such an offer, in accord
with subsection 6b.
d) Convey the object code by offering access from a designated
place (gratis or for a charge), and offer equivalent access to the
Corresponding Source in the same way through the same place at no
further charge. You need not require recipients to copy the
Corresponding Source along with the object code. If the place to
copy the object code is a network server, the Corresponding Source
may be on a different server (operated by you or a third party)
that supports equivalent copying facilities, provided you maintain
clear directions next to the object code saying where to find the
Corresponding Source. Regardless of what server hosts the
Corresponding Source, you remain obligated to ensure that it is
available for as long as needed to satisfy these requirements.
e) Convey the object code using peer-to-peer transmission, provided
you inform other peers where the object code and Corresponding
Source of the work are being offered to the general public at no
charge under subsection 6d.
A separable portion of the object code, whose source code is excluded
from the Corresponding Source as a System Library, need not be
included in conveying the object code work.
A "User Product" is either (1) a "consumer product", which means any
tangible personal property which is normally used for personal, family,
or household purposes, or (2) anything designed or sold for incorporation
into a dwelling. In determining whether a product is a consumer product,
doubtful cases shall be resolved in favor of coverage. For a particular
product received by a particular user, "normally used" refers to a
typical or common use of that class of product, regardless of the status
of the particular user or of the way in which the particular user
actually uses, or expects or is expected to use, the product. A product
is a consumer product regardless of whether the product has substantial
commercial, industrial or non-consumer uses, unless such uses represent
the only significant mode of use of the product.
"Installation Information" for a User Product means any methods,
procedures, authorization keys, or other information required to install
and execute modified versions of a covered work in that User Product from
a modified version of its Corresponding Source. The information must
suffice to ensure that the continued functioning of the modified object
code is in no case prevented or interfered with solely because
modification has been made.
If you convey an object code work under this section in, or with, or
specifically for use in, a User Product, and the conveying occurs as
part of a transaction in which the right of possession and use of the
User Product is transferred to the recipient in perpetuity or for a
fixed term (regardless of how the transaction is characterized), the
Corresponding Source conveyed under this section must be accompanied
by the Installation Information. But this requirement does not apply
if neither you nor any third party retains the ability to install
modified object code on the User Product (for example, the work has
been installed in ROM).
The requirement to provide Installation Information does not include a
requirement to continue to provide support service, warranty, or updates
for a work that has been modified or installed by the recipient, or for
the User Product in which it has been modified or installed. Access to a
network may be denied when the modification itself materially and
adversely affects the operation of the network or violates the rules and
protocols for communication across the network.
Corresponding Source conveyed, and Installation Information provided,
in accord with this section must be in a format that is publicly
documented (and with an implementation available to the public in
source code form), and must require no special password or key for
unpacking, reading or copying.
7. Additional Terms.
"Additional permissions" are terms that supplement the terms of this
License by making exceptions from one or more of its conditions.
Additional permissions that are applicable to the entire Program shall
be treated as though they were included in this License, to the extent
that they are valid under applicable law. If additional permissions
apply only to part of the Program, that part may be used separately
under those permissions, but the entire Program remains governed by
this License without regard to the additional permissions.
When you convey a copy of a covered work, you may at your option
remove any additional permissions from that copy, or from any part of
it. (Additional permissions may be written to require their own
removal in certain cases when you modify the work.) You may place
additional permissions on material, added by you to a covered work,
for which you have or can give appropriate copyright permission.
Notwithstanding any other provision of this License, for material you
add to a covered work, you may (if authorized by the copyright holders of
that material) supplement the terms of this License with terms:
a) Disclaiming warranty or limiting liability differently from the
terms of sections 15 and 16 of this License; or
b) Requiring preservation of specified reasonable legal notices or
author attributions in that material or in the Appropriate Legal
Notices displayed by works containing it; or
c) Prohibiting misrepresentation of the origin of that material, or
requiring that modified versions of such material be marked in
reasonable ways as different from the original version; or
d) Limiting the use for publicity purposes of names of licensors or
authors of the material; or
e) Declining to grant rights under trademark law for use of some
trade names, trademarks, or service marks; or
f) Requiring indemnification of licensors and authors of that
material by anyone who conveys the material (or modified versions of
it) with contractual assumptions of liability to the recipient, for
any liability that these contractual assumptions directly impose on
those licensors and authors.
All other non-permissive additional terms are considered "further
restrictions" within the meaning of section 10. If the Program as you
received it, or any part of it, contains a notice stating that it is
governed by this License along with a term that is a further
restriction, you may remove that term. If a license document contains
a further restriction but permits relicensing or conveying under this
License, you may add to a covered work material governed by the terms
of that license document, provided that the further restriction does
not survive such relicensing or conveying.
If you add terms to a covered work in accord with this section, you
must place, in the relevant source files, a statement of the
additional terms that apply to those files, or a notice indicating
where to find the applicable terms.
Additional terms, permissive or non-permissive, may be stated in the
form of a separately written license, or stated as exceptions;
the above requirements apply either way.
8. Termination.
You may not propagate or modify a covered work except as expressly
provided under this License. Any attempt otherwise to propagate or
modify it is void, and will automatically terminate your rights under
this License (including any patent licenses granted under the third
paragraph of section 11).
However, if you cease all violation of this License, then your
license from a particular copyright holder is reinstated (a)
provisionally, unless and until the copyright holder explicitly and
finally terminates your license, and (b) permanently, if the copyright
holder fails to notify you of the violation by some reasonable means
prior to 60 days after the cessation.
Moreover, your license from a particular copyright holder is
reinstated permanently if the copyright holder notifies you of the
violation by some reasonable means, this is the first time you have
received notice of violation of this License (for any work) from that
copyright holder, and you cure the violation prior to 30 days after
your receipt of the notice.
Termination of your rights under this section does not terminate the
licenses of parties who have received copies or rights from you under
this License. If your rights have been terminated and not permanently
reinstated, you do not qualify to receive new licenses for the same
material under section 10.
9. Acceptance Not Required for Having Copies.
You are not required to accept this License in order to receive or
run a copy of the Program. Ancillary propagation of a covered work
occurring solely as a consequence of using peer-to-peer transmission
to receive a copy likewise does not require acceptance. However,
nothing other than this License grants you permission to propagate or
modify any covered work. These actions infringe copyright if you do
not accept this License. Therefore, by modifying or propagating a
covered work, you indicate your acceptance of this License to do so.
10. Automatic Licensing of Downstream Recipients.
Each time you convey a covered work, the recipient automatically
receives a license from the original licensors, to run, modify and
propagate that work, subject to this License. You are not responsible
for enforcing compliance by third parties with this License.
An "entity transaction" is a transaction transferring control of an
organization, or substantially all assets of one, or subdividing an
organization, or merging organizations. If propagation of a covered
work results from an entity transaction, each party to that
transaction who receives a copy of the work also receives whatever
licenses to the work the party's predecessor in interest had or could
give under the previous paragraph, plus a right to possession of the
Corresponding Source of the work from the predecessor in interest, if
the predecessor has it or can get it with reasonable efforts.
You may not impose any further restrictions on the exercise of the
rights granted or affirmed under this License. For example, you may
not impose a license fee, royalty, or other charge for exercise of
rights granted under this License, and you may not initiate litigation
(including a cross-claim or counterclaim in a lawsuit) alleging that
any patent claim is infringed by making, using, selling, offering for
sale, or importing the Program or any portion of it.
11. Patents.
A "contributor" is a copyright holder who authorizes use under this
License of the Program or a work on which the Program is based. The
work thus licensed is called the contributor's "contributor version".
A contributor's "essential patent claims" are all patent claims
owned or controlled by the contributor, whether already acquired or
hereafter acquired, that would be infringed by some manner, permitted
by this License, of making, using, or selling its contributor version,
but do not include claims that would be infringed only as a
consequence of further modification of the contributor version. For
purposes of this definition, "control" includes the right to grant
patent sublicenses in a manner consistent with the requirements of
this License.
Each contributor grants you a non-exclusive, worldwide, royalty-free
patent license under the contributor's essential patent claims, to
make, use, sell, offer for sale, import and otherwise run, modify and
propagate the contents of its contributor version.
In the following three paragraphs, a "patent license" is any express
agreement or commitment, however denominated, not to enforce a patent
(such as an express permission to practice a patent or covenant not to
sue for patent infringement). To "grant" such a patent license to a
party means to make such an agreement or commitment not to enforce a
patent against the party.
If you convey a covered work, knowingly relying on a patent license,
and the Corresponding Source of the work is not available for anyone
to copy, free of charge and under the terms of this License, through a
publicly available network server or other readily accessible means,
then you must either (1) cause the Corresponding Source to be so
available, or (2) arrange to deprive yourself of the benefit of the
patent license for this particular work, or (3) arrange, in a manner
consistent with the requirements of this License, to extend the patent
license to downstream recipients. "Knowingly relying" means you have
actual knowledge that, but for the patent license, your conveying the
covered work in a country, or your recipient's use of the covered work
in a country, would infringe one or more identifiable patents in that
country that you have reason to believe are valid.
If, pursuant to or in connection with a single transaction or
arrangement, you convey, or propagate by procuring conveyance of, a
covered work, and grant a patent license to some of the parties
receiving the covered work authorizing them to use, propagate, modify
or convey a specific copy of the covered work, then the patent license
you grant is automatically extended to all recipients of the covered
work and works based on it.
A patent license is "discriminatory" if it does not include within
the scope of its coverage, prohibits the exercise of, or is
conditioned on the non-exercise of one or more of the rights that are
specifically granted under this License. You may not convey a covered
work if you are a party to an arrangement with a third party that is
in the business of distributing software, under which you make payment
to the third party based on the extent of your activity of conveying
the work, and under which the third party grants, to any of the
parties who would receive the covered work from you, a discriminatory
patent license (a) in connection with copies of the covered work
conveyed by you (or copies made from those copies), or (b) primarily
for and in connection with specific products or compilations that
contain the covered work, unless you entered into that arrangement,
or that patent license was granted, prior to 28 March 2007.
Nothing in this License shall be construed as excluding or limiting
any implied license or other defenses to infringement that may
otherwise be available to you under applicable patent law.
12. No Surrender of Others' Freedom.
If conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot convey a
covered work so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you may
not convey it at all. For example, if you agree to terms that obligate you
to collect a royalty for further conveying from those to whom you convey
the Program, the only way you could satisfy both those terms and this
License would be to refrain entirely from conveying the Program.
13. Use with the GNU Affero General Public License.
Notwithstanding any other provision of this License, you have
permission to link or combine any covered work with a work licensed
under version 3 of the GNU Affero General Public License into a single
combined work, and to convey the resulting work. The terms of this
License will continue to apply to the part which is the covered work,
but the special requirements of the GNU Affero General Public License,
section 13, concerning interaction through a network will apply to the
combination as such.
14. Revised Versions of this License.
The Free Software Foundation may publish revised and/or new versions of
the GNU General Public License from time to time. Such new versions will
be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the
Program specifies that a certain numbered version of the GNU General
Public License "or any later version" applies to it, you have the
option of following the terms and conditions either of that numbered
version or of any later version published by the Free Software
Foundation. If the Program does not specify a version number of the
GNU General Public License, you may choose any version ever published
by the Free Software Foundation.
If the Program specifies that a proxy can decide which future
versions of the GNU General Public License can be used, that proxy's
public statement of acceptance of a version permanently authorizes you
to choose that version for the Program.
Later license versions may give you additional or different
permissions. However, no additional obligations are imposed on any
author or copyright holder as a result of your choosing to follow a
later version.
15. Disclaimer of Warranty.
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
16. Limitation of Liability.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES.
17. Interpretation of Sections 15 and 16.
If the disclaimer of warranty and limitation of liability provided
above cannot be given local legal effect according to their terms,
reviewing courts shall apply local law that most closely approximates
an absolute waiver of all civil liability in connection with the
Program, unless a warranty or assumption of liability accompanies a
copy of the Program in return for a fee.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
state the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
Also add information on how to contact you by electronic and paper mail.
If the program does terminal interaction, make it output a short
notice like this when it starts in an interactive mode:
<program> Copyright (C) <year> <name of author>
This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
This is free software, and you are welcome to redistribute it
under certain conditions; type `show c' for details.
The hypothetical commands `show w' and `show c' should show the appropriate
parts of the General Public License. Of course, your program's commands
might be different; for a GUI interface, you would use an "about box".
You should also get your employer (if you work as a programmer) or school,
if any, to sign a "copyright disclaimer" for the program, if necessary.
For more information on this, and how to apply and follow the GNU GPL, see
<http://www.gnu.org/licenses/>.
The GNU General Public License does not permit incorporating your program
into proprietary programs. If your program is a subroutine library, you
may consider it more useful to permit linking proprietary applications with
the library. If this is what you want to do, use the GNU Lesser General
Public License instead of this License. But first, please read
<http://www.gnu.org/philosophy/why-not-lgpl.html>.
================================================
FILE: README.md
================================================
# Yolov5 Face Detection
## Description
The project is a wrap over [yolov5-face](https://github.com/deepcam-cn/yolov5-face) repo. Made simple portable interface for model import and inference. Model detects faces on images and returns bounding boxes and coordinates of 5 facial keypoints, which can be used for face alignment.
## Installation
```bash
pip install -r requirements.txt
```
## Usage example
```python
from face_detector import YoloDetector
import numpy as np
from PIL import Image
model = YoloDetector(target_size=720, device="cuda:0", min_face=90)
orgimg = np.array(Image.open('test_image.jpg'))
bboxes,points = model.predict(orgimg)
```
You can also pass several images packed in a list to get multi-image predictions:
```python
bboxes,points = model.predict([image1,image2])
```
You can align faces, using `align` class method for predicted keypoints. May be useful in conjunction with facial recognition neural network to increase accuracy:
```python
crops = model.align(orgimg, points[0])
```
If you want to use model class outside root folder, export it into you PYTHONPATH:
```bash
export PYTHONPATH="${PYTHONPATH}:/path/to/yoloface/project/"
```
or the same from python:
```python
import sys
sys.path.append("/path/to/yoloface/project/")
```
## Other pretrained models
You can use any model from [yolov5-face](https://github.com/deepcam-cn/yolov5-face#pretrained-models) repo. Default models are saved as entire torch module and are bound to the specific classes and the exact directory structure used when the model was saved by authors. To make model portable and run it via my interface you must save it as pytorch state_dict and put new weights in `weights/` folder. Example below:
```python
model = torch.load('weights/yolov5m-face.pt', map_location='cpu')['model']
torch.save(model.state_dict(),'path/to/project/weights/yolov5m_state_dict.pt')
```
Then when creating YoloDetector class object, pass new model name and corresponding yaml config from `models/` folder as class arguments.
Example below:
```python
model = YoloFace(weights_name='yolov5m_state_dict.pt',config_name='yolov5m.yaml',target_size=720)
```
## Result example
<img src="/results/result_example.jpg" width="600"/>
## Citiation
Thanks [deepcam-cn](https://github.com/deepcam-cn/yolov5-face) for pretrained models.
================================================
FILE: face_detector.py
================================================
import joblib
import os
import sys
import torch
import torch.nn as nn
import numpy as np
import cv2
import copy
import scipy
import pathlib
import warnings
from math import sqrt
sys.path.append(os.path.abspath(os.path.join(os.path.dirname("__file__"), '..')))
from models.common import Conv
from models.yolo import Model
from utils.datasets import letterbox
from utils.preprocess_utils import align_faces
from utils.general import check_img_size, non_max_suppression_face, \
scale_coords,scale_coords_landmarks,filter_boxes
class YoloDetector:
def __init__(self, weights_name='yolov5n_state_dict.pt', config_name='yolov5n.yaml', device='cuda:0', min_face=100, target_size=None, frontal=False):
"""
weights_name: name of file with network weights in weights/ folder.
config_name: name of .yaml config with network configuration from models/ folder.
device : pytorch device. Use 'cuda:0', 'cuda:1', e.t.c to use gpu or 'cpu' to use cpu.
min_face : minimal face size in pixels.
target_size : target size of smaller image axis (choose lower for faster work). e.g. 480, 720, 1080. Choose None for original resolution.
frontal : if True tries to filter nonfrontal faces by keypoints location. CURRENTRLY UNSUPPORTED.
"""
self._class_path = pathlib.Path(__file__).parent.absolute()#os.path.dirname(inspect.getfile(self.__class__))
self.device = device
self.target_size = target_size
self.min_face = min_face
self.frontal = frontal
if self.frontal:
print('Currently unavailable')
# self.anti_profile = joblib.load(os.path.join(self._class_path, 'models/anti_profile/anti_profile_xgb_new.pkl'))
self.detector = self.init_detector(weights_name,config_name)
def init_detector(self,weights_name,config_name):
print(self.device)
model_path = os.path.join(self._class_path,'weights/',weights_name)
print(model_path)
config_path = os.path.join(self._class_path,'models/',config_name)
state_dict = torch.load(model_path)
detector = Model(cfg=config_path)
detector.load_state_dict(state_dict)
detector = detector.to(self.device).float().eval()
for m in detector.modules():
if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]:
m.inplace = True # pytorch 1.7.0 compatibility
elif type(m) is Conv:
m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility
return detector
def _preprocess(self,imgs):
"""
Preprocessing image before passing through the network. Resize and conversion to torch tensor.
"""
pp_imgs = []
for img in imgs:
h0, w0 = img.shape[:2] # orig hw
if self.target_size:
r = self.target_size / min(h0, w0) # resize image to img_size
if r < 1:
img = cv2.resize(img, (int(w0 * r), int(h0 * r)), interpolation=cv2.INTER_LINEAR)
imgsz = check_img_size(max(img.shape[:2]), s=self.detector.stride.max()) # check img_size
img = letterbox(img, new_shape=imgsz)[0]
pp_imgs.append(img)
pp_imgs = np.array(pp_imgs)
pp_imgs = pp_imgs.transpose(0, 3, 1, 2)
pp_imgs = torch.from_numpy(pp_imgs).to(self.device)
pp_imgs = pp_imgs.float() # uint8 to fp16/32
pp_imgs /= 255.0 # 0 - 255 to 0.0 - 1.0
return pp_imgs
def _postprocess(self, imgs, origimgs, pred, conf_thres, iou_thres):
"""
Postprocessing of raw pytorch model output.
Returns:
bboxes: list of arrays with 4 coordinates of bounding boxes with format x1,y1,x2,y2.
points: list of arrays with coordinates of 5 facial keypoints (eyes, nose, lips corners).
"""
bboxes = [[] for i in range(len(origimgs))]
landmarks = [[] for i in range(len(origimgs))]
pred = non_max_suppression_face(pred, conf_thres, iou_thres)
for i in range(len(origimgs)):
img_shape = origimgs[i].shape
h,w = img_shape[:2]
gn = torch.tensor(img_shape)[[1, 0, 1, 0]] # normalization gain whwh
gn_lks = torch.tensor(img_shape)[[1, 0, 1, 0, 1, 0, 1, 0, 1, 0]] # normalization gain landmarks
det = pred[i].cpu()
scaled_bboxes = scale_coords(imgs[i].shape[1:], det[:, :4], img_shape).round()
scaled_cords = scale_coords_landmarks(imgs[i].shape[1:], det[:, 5:15], img_shape).round()
for j in range(det.size()[0]):
box = (det[j, :4].view(1, 4) / gn).view(-1).tolist()
box = list(map(int,[box[0]*w,box[1]*h,box[2]*w,box[3]*h]))
if box[3] - box[1] < self.min_face:
continue
lm = (det[j, 5:15].view(1, 10) / gn_lks).view(-1).tolist()
lm = list(map(int,[i*w if j%2==0 else i*h for j,i in enumerate(lm)]))
lm = [lm[i:i+2] for i in range(0,len(lm),2)]
bboxes[i].append(box)
landmarks[i].append(lm)
return bboxes, landmarks
def get_frontal_predict(self, box, points):
'''
Make a decision whether face is frontal by keypoints.
Returns:
True if face is frontal, False otherwise.
'''
cur_points = points.astype('int')
x1, y1, x2, y2 = box[0:4]
w = x2-x1
h = y2-y1
diag = sqrt(w**2+h**2)
dist = scipy.spatial.distance.pdist(cur_points)/diag
predict = self.anti_profile.predict(dist.reshape(1, -1))[0]
if predict == 0:
return True
else:
return False
def align(self, img, points):
'''
Align faces, found on images.
Params:
img: Single image, used in predict method.
points: list of keypoints, produced in predict method.
Returns:
crops: list of croped and aligned faces of shape (112,112,3).
'''
crops = [align_faces(img,landmark=np.array(i)) for i in points]
return crops
def predict(self, imgs, conf_thres = 0.3, iou_thres = 0.5):
'''
Get bbox coordinates and keypoints of faces on original image.
Params:
imgs: image or list of images to detect faces on
conf_thres: confidence threshold for each prediction
iou_thres: threshold for NMS (filtering of intersecting bboxes)
Returns:
bboxes: list of arrays with 4 coordinates of bounding boxes with format x1,y1,x2,y2.
points: list of arrays with coordinates of 5 facial keypoints (eyes, nose, lips corners).
'''
one_by_one = False
# Pass input images through face detector
if type(imgs) != list:
images = [imgs]
else:
images = imgs
one_by_one = False
shapes = {arr.shape for arr in images}
if len(shapes) != 1:
one_by_one = True
warnings.warn(f"Can't use batch predict due to different shapes of input images. Using one by one strategy.")
origimgs = copy.deepcopy(images)
if one_by_one:
images = [self._preprocess([img]) for img in images]
bboxes = [[] for i in range(len(origimgs))]
points = [[] for i in range(len(origimgs))]
for num, img in enumerate(images):
with torch.inference_mode(): # change this with torch.no_grad() for pytorch <1.8 compatibility
single_pred = self.detector(img)[0]
print(single_pred.shape)
bb, pt = self._postprocess(img, [origimgs[num]], single_pred, conf_thres, iou_thres)
#print(bb)
bboxes[num] = bb[0]
points[num] = pt[0]
else:
images = self._preprocess(images)
with torch.inference_mode(): # change this with torch.no_grad() for pytorch <1.8 compatibility
pred = self.detector(images)[0]
bboxes, points = self._postprocess(images, origimgs, pred, conf_thres, iou_thres)
return bboxes, points
def __call__(self,*args):
return self.predict(*args)
if __name__=='__main__':
a = YoloDetector()
================================================
FILE: models/__init__.py
================================================
================================================
FILE: models/common.py
================================================
# This file contains modules common to various models
import math
import numpy as np
import requests
import torch
import torch.nn as nn
from PIL import Image, ImageDraw
from utils.datasets import letterbox
from utils.general import non_max_suppression, make_divisible, scale_coords, xyxy2xywh
from utils.plots import color_list
def autopad(k, p=None): # kernel, padding
# Pad to 'same'
if p is None:
p = k // 2 if isinstance(k, int) else [x // 2 for x in k] # auto-pad
return p
def channel_shuffle(x, groups):
batchsize, num_channels, height, width = x.data.size()
channels_per_group = num_channels // groups
# reshape
x = x.view(batchsize, groups, channels_per_group, height, width)
x = torch.transpose(x, 1, 2).contiguous()
# flatten
x = x.view(batchsize, -1, height, width)
return x
def DWConv(c1, c2, k=1, s=1, act=True):
# Depthwise convolution
return Conv(c1, c2, k, s, g=math.gcd(c1, c2), act=act)
class Conv(nn.Module):
# Standard convolution
def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups
super(Conv, self).__init__()
self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
self.bn = nn.BatchNorm2d(c2)
self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
#self.act = self.act = nn.LeakyReLU(0.1, inplace=True) if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
def forward(self, x):
return self.act(self.bn(self.conv(x)))
def fuseforward(self, x):
return self.act(self.conv(x))
class StemBlock(nn.Module):
def __init__(self, c1, c2, k=3, s=2, p=None, g=1, act=True):
super(StemBlock, self).__init__()
self.stem_1 = Conv(c1, c2, k, s, p, g, act)
self.stem_2a = Conv(c2, c2 // 2, 1, 1, 0)
self.stem_2b = Conv(c2 // 2, c2, 3, 2, 1)
self.stem_2p = nn.MaxPool2d(kernel_size=2,stride=2,ceil_mode=True)
self.stem_3 = Conv(c2 * 2, c2, 1, 1, 0)
def forward(self, x):
stem_1_out = self.stem_1(x)
stem_2a_out = self.stem_2a(stem_1_out)
stem_2b_out = self.stem_2b(stem_2a_out)
stem_2p_out = self.stem_2p(stem_1_out)
out = self.stem_3(torch.cat((stem_2b_out,stem_2p_out),1))
return out
class Bottleneck(nn.Module):
# Standard bottleneck
def __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_out, shortcut, groups, expansion
super(Bottleneck, self).__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c_, c2, 3, 1, g=g)
self.add = shortcut and c1 == c2
def forward(self, x):
return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
class BottleneckCSP(nn.Module):
# CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks
def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion
super(BottleneckCSP, self).__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False)
self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
self.cv4 = Conv(2 * c_, c2, 1, 1)
self.bn = nn.BatchNorm2d(2 * c_) # applied to cat(cv2, cv3)
self.act = nn.LeakyReLU(0.1, inplace=True)
self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
def forward(self, x):
y1 = self.cv3(self.m(self.cv1(x)))
y2 = self.cv2(x)
return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1))))
class C3(nn.Module):
# CSP Bottleneck with 3 convolutions
def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion
super(C3, self).__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c1, c_, 1, 1)
self.cv3 = Conv(2 * c_, c2, 1) # act=FReLU(c2)
self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
def forward(self, x):
return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))
class ShuffleV2Block(nn.Module):
def __init__(self, inp, oup, stride):
super(ShuffleV2Block, self).__init__()
if not (1 <= stride <= 3):
raise ValueError('illegal stride value')
self.stride = stride
branch_features = oup // 2
assert (self.stride != 1) or (inp == branch_features << 1)
if self.stride > 1:
self.branch1 = nn.Sequential(
self.depthwise_conv(inp, inp, kernel_size=3, stride=self.stride, padding=1),
nn.BatchNorm2d(inp),
nn.Conv2d(inp, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.SiLU(),
)
else:
self.branch1 = nn.Sequential()
self.branch2 = nn.Sequential(
nn.Conv2d(inp if (self.stride > 1) else branch_features, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.SiLU(),
self.depthwise_conv(branch_features, branch_features, kernel_size=3, stride=self.stride, padding=1),
nn.BatchNorm2d(branch_features),
nn.Conv2d(branch_features, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.SiLU(),
)
@staticmethod
def depthwise_conv(i, o, kernel_size, stride=1, padding=0, bias=False):
return nn.Conv2d(i, o, kernel_size, stride, padding, bias=bias, groups=i)
def forward(self, x):
if self.stride == 1:
x1, x2 = x.chunk(2, dim=1)
out = torch.cat((x1, self.branch2(x2)), dim=1)
else:
out = torch.cat((self.branch1(x), self.branch2(x)), dim=1)
out = channel_shuffle(out, 2)
return out
class SPP(nn.Module):
# Spatial pyramid pooling layer used in YOLOv3-SPP
def __init__(self, c1, c2, k=(5, 9, 13)):
super(SPP, self).__init__()
c_ = c1 // 2 # hidden channels
self.cv1 = Conv(c1, c_, 1, 1)
self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1)
self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k])
def forward(self, x):
x = self.cv1(x)
return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))
class Focus(nn.Module):
# Focus wh information into c-space
def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups
super(Focus, self).__init__()
self.conv = Conv(c1 * 4, c2, k, s, p, g, act)
# self.contract = Contract(gain=2)
def forward(self, x): # x(b,c,w,h) -> y(b,4c,w/2,h/2)
return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1))
# return self.conv(self.contract(x))
class Contract(nn.Module):
# Contract width-height into channels, i.e. x(1,64,80,80) to x(1,256,40,40)
def __init__(self, gain=2):
super().__init__()
self.gain = gain
def forward(self, x):
N, C, H, W = x.size() # assert (H / s == 0) and (W / s == 0), 'Indivisible gain'
s = self.gain
x = x.view(N, C, H // s, s, W // s, s) # x(1,64,40,2,40,2)
x = x.permute(0, 3, 5, 1, 2, 4).contiguous() # x(1,2,2,64,40,40)
return x.view(N, C * s * s, H // s, W // s) # x(1,256,40,40)
class Expand(nn.Module):
# Expand channels into width-height, i.e. x(1,64,80,80) to x(1,16,160,160)
def __init__(self, gain=2):
super().__init__()
self.gain = gain
def forward(self, x):
N, C, H, W = x.size() # assert C / s ** 2 == 0, 'Indivisible gain'
s = self.gain
x = x.view(N, s, s, C // s ** 2, H, W) # x(1,2,2,16,80,80)
x = x.permute(0, 3, 4, 1, 5, 2).contiguous() # x(1,16,80,2,80,2)
return x.view(N, C // s ** 2, H * s, W * s) # x(1,16,160,160)
class Concat(nn.Module):
# Concatenate a list of tensors along dimension
def __init__(self, dimension=1):
super(Concat, self).__init__()
self.d = dimension
def forward(self, x):
return torch.cat(x, self.d)
class NMS(nn.Module):
# Non-Maximum Suppression (NMS) module
conf = 0.25 # confidence threshold
iou = 0.45 # IoU threshold
classes = None # (optional list) filter by class
def __init__(self):
super(NMS, self).__init__()
def forward(self, x):
return non_max_suppression(x[0], conf_thres=self.conf, iou_thres=self.iou, classes=self.classes)
class autoShape(nn.Module):
# input-robust model wrapper for passing cv2/np/PIL/torch inputs. Includes preprocessing, inference and NMS
img_size = 640 # inference size (pixels)
conf = 0.25 # NMS confidence threshold
iou = 0.45 # NMS IoU threshold
classes = None # (optional list) filter by class
def __init__(self, model):
super(autoShape, self).__init__()
self.model = model.eval()
def autoshape(self):
print('autoShape already enabled, skipping... ') # model already converted to model.autoshape()
return self
def forward(self, imgs, size=640, augment=False, profile=False):
# Inference from various sources. For height=720, width=1280, RGB images example inputs are:
# filename: imgs = 'data/samples/zidane.jpg'
# URI: = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/zidane.jpg'
# OpenCV: = cv2.imread('image.jpg')[:,:,::-1] # HWC BGR to RGB x(720,1280,3)
# PIL: = Image.open('image.jpg') # HWC x(720,1280,3)
# numpy: = np.zeros((720,1280,3)) # HWC
# torch: = torch.zeros(16,3,720,1280) # BCHW
# multiple: = [Image.open('image1.jpg'), Image.open('image2.jpg'), ...] # list of images
p = next(self.model.parameters()) # for device and type
if isinstance(imgs, torch.Tensor): # torch
return self.model(imgs.to(p.device).type_as(p), augment, profile) # inference
# Pre-process
n, imgs = (len(imgs), imgs) if isinstance(imgs, list) else (1, [imgs]) # number of images, list of images
shape0, shape1 = [], [] # image and inference shapes
for i, im in enumerate(imgs):
if isinstance(im, str): # filename or uri
im = Image.open(requests.get(im, stream=True).raw if im.startswith('http') else im) # open
im = np.array(im) # to numpy
if im.shape[0] < 5: # image in CHW
im = im.transpose((1, 2, 0)) # reverse dataloader .transpose(2, 0, 1)
im = im[:, :, :3] if im.ndim == 3 else np.tile(im[:, :, None], 3) # enforce 3ch input
s = im.shape[:2] # HWC
shape0.append(s) # image shape
g = (size / max(s)) # gain
shape1.append([y * g for y in s])
imgs[i] = im # update
shape1 = [make_divisible(x, int(self.stride.max())) for x in np.stack(shape1, 0).max(0)] # inference shape
x = [letterbox(im, new_shape=shape1, auto=False)[0] for im in imgs] # pad
x = np.stack(x, 0) if n > 1 else x[0][None] # stack
x = np.ascontiguousarray(x.transpose((0, 3, 1, 2))) # BHWC to BCHW
x = torch.from_numpy(x).to(p.device).type_as(p) / 255. # uint8 to fp16/32
# Inference
with torch.no_grad():
y = self.model(x, augment, profile)[0] # forward
y = non_max_suppression(y, conf_thres=self.conf, iou_thres=self.iou, classes=self.classes) # NMS
# Post-process
for i in range(n):
scale_coords(shape1, y[i][:, :4], shape0[i])
return Detections(imgs, y, self.names)
class Detections:
# detections class for YOLOv5 inference results
def __init__(self, imgs, pred, names=None):
super(Detections, self).__init__()
d = pred[0].device # device
gn = [torch.tensor([*[im.shape[i] for i in [1, 0, 1, 0]], 1., 1.], device=d) for im in imgs] # normalizations
self.imgs = imgs # list of images as numpy arrays
self.pred = pred # list of tensors pred[0] = (xyxy, conf, cls)
self.names = names # class names
self.xyxy = pred # xyxy pixels
self.xywh = [xyxy2xywh(x) for x in pred] # xywh pixels
self.xyxyn = [x / g for x, g in zip(self.xyxy, gn)] # xyxy normalized
self.xywhn = [x / g for x, g in zip(self.xywh, gn)] # xywh normalized
self.n = len(self.pred)
def display(self, pprint=False, show=False, save=False, render=False):
colors = color_list()
for i, (img, pred) in enumerate(zip(self.imgs, self.pred)):
str = f'Image {i + 1}/{len(self.pred)}: {img.shape[0]}x{img.shape[1]} '
if pred is not None:
for c in pred[:, -1].unique():
n = (pred[:, -1] == c).sum() # detections per class
str += f'{n} {self.names[int(c)]}s, ' # add to string
if show or save or render:
img = Image.fromarray(img.astype(np.uint8)) if isinstance(img, np.ndarray) else img # from np
for *box, conf, cls in pred: # xyxy, confidence, class
# str += '%s %.2f, ' % (names[int(cls)], conf) # label
ImageDraw.Draw(img).rectangle(box, width=4, outline=colors[int(cls) % 10]) # plot
if pprint:
print(str)
if show:
img.show(f'Image {i}') # show
if save:
f = f'results{i}.jpg'
str += f"saved to '{f}'"
img.save(f) # save
if render:
self.imgs[i] = np.asarray(img)
def print(self):
self.display(pprint=True) # print results
def show(self):
self.display(show=True) # show results
def save(self):
self.display(save=True) # save results
def render(self):
self.display(render=True) # render results
return self.imgs
def __len__(self):
return self.n
def tolist(self):
# return a list of Detections objects, i.e. 'for result in results.tolist():'
x = [Detections([self.imgs[i]], [self.pred[i]], self.names) for i in range(self.n)]
for d in x:
for k in ['imgs', 'pred', 'xyxy', 'xyxyn', 'xywh', 'xywhn']:
setattr(d, k, getattr(d, k)[0]) # pop out of list
return x
class Classify(nn.Module):
# Classification head, i.e. x(b,c1,20,20) to x(b,c2)
def __init__(self, c1, c2, k=1, s=1, p=None, g=1): # ch_in, ch_out, kernel, stride, padding, groups
super(Classify, self).__init__()
self.aap = nn.AdaptiveAvgPool2d(1) # to x(b,c1,1,1)
self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g) # to x(b,c2,1,1)
self.flat = nn.Flatten()
def forward(self, x):
z = torch.cat([self.aap(y) for y in (x if isinstance(x, list) else [x])], 1) # cat if list
return self.flat(self.conv(z)) # flatten to x(b,c2)
================================================
FILE: models/experimental.py
================================================
# This file contains experimental modules
import numpy as np
import torch
import torch.nn as nn
from models.common import Conv, DWConv
from utils.google_utils import attempt_download
class CrossConv(nn.Module):
# Cross Convolution Downsample
def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False):
# ch_in, ch_out, kernel, stride, groups, expansion, shortcut
super(CrossConv, self).__init__()
c_ = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, c_, (1, k), (1, s))
self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g)
self.add = shortcut and c1 == c2
def forward(self, x):
return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
class Sum(nn.Module):
# Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070
def __init__(self, n, weight=False): # n: number of inputs
super(Sum, self).__init__()
self.weight = weight # apply weights boolean
self.iter = range(n - 1) # iter object
if weight:
self.w = nn.Parameter(-torch.arange(1., n) / 2, requires_grad=True) # layer weights
def forward(self, x):
y = x[0] # no weight
if self.weight:
w = torch.sigmoid(self.w) * 2
for i in self.iter:
y = y + x[i + 1] * w[i]
else:
for i in self.iter:
y = y + x[i + 1]
return y
class GhostConv(nn.Module):
# Ghost Convolution https://github.com/huawei-noah/ghostnet
def __init__(self, c1, c2, k=1, s=1, g=1, act=True): # ch_in, ch_out, kernel, stride, groups
super(GhostConv, self).__init__()
c_ = c2 // 2 # hidden channels
self.cv1 = Conv(c1, c_, k, s, None, g, act)
self.cv2 = Conv(c_, c_, 5, 1, None, c_, act)
def forward(self, x):
y = self.cv1(x)
return torch.cat([y, self.cv2(y)], 1)
class GhostBottleneck(nn.Module):
# Ghost Bottleneck https://github.com/huawei-noah/ghostnet
def __init__(self, c1, c2, k, s):
super(GhostBottleneck, self).__init__()
c_ = c2 // 2
self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1), # pw
DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(), # dw
GhostConv(c_, c2, 1, 1, act=False)) # pw-linear
self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False),
Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity()
def forward(self, x):
return self.conv(x) + self.shortcut(x)
class MixConv2d(nn.Module):
# Mixed Depthwise Conv https://arxiv.org/abs/1907.09595
def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):
super(MixConv2d, self).__init__()
groups = len(k)
if equal_ch: # equal c_ per group
i = torch.linspace(0, groups - 1E-6, c2).floor() # c2 indices
c_ = [(i == g).sum() for g in range(groups)] # intermediate channels
else: # equal weight.numel() per group
b = [c2] + [0] * groups
a = np.eye(groups + 1, groups, k=-1)
a -= np.roll(a, 1, axis=1)
a *= np.array(k) ** 2
a[0] = 1
c_ = np.linalg.lstsq(a, b, rcond=None)[0].round() # solve for equal weight indices, ax = b
self.m = nn.ModuleList([nn.Conv2d(c1, int(c_[g]), k[g], s, k[g] // 2, bias=False) for g in range(groups)])
self.bn = nn.BatchNorm2d(c2)
self.act = nn.LeakyReLU(0.1, inplace=True)
def forward(self, x):
return x + self.act(self.bn(torch.cat([m(x) for m in self.m], 1)))
class Ensemble(nn.ModuleList):
# Ensemble of models
def __init__(self):
super(Ensemble, self).__init__()
def forward(self, x, augment=False):
y = []
for module in self:
y.append(module(x, augment)[0])
# y = torch.stack(y).max(0)[0] # max ensemble
# y = torch.stack(y).mean(0) # mean ensemble
y = torch.cat(y, 1) # nms ensemble
return y, None # inference, train output
def attempt_load(weights, map_location=None):
# Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a
model = Ensemble()
for w in weights if isinstance(weights, list) else [weights]:
attempt_download(w)
model.append(torch.load(w, map_location=map_location)['model'].float().fuse().eval()) # load FP32 model
# Compatibility updates
for m in model.modules():
if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]:
m.inplace = True # pytorch 1.7.0 compatibility
elif type(m) is Conv:
m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility
if len(model) == 1:
return model[-1] # return model
else:
print('Ensemble created with %s\n' % weights)
for k in ['names', 'stride']:
setattr(model, k, getattr(model[-1], k))
return model # return ensemble
================================================
FILE: models/export.py
================================================
"""Exports a YOLOv5 *.pt model to ONNX and TorchScript formats
Usage:
$ export PYTHONPATH="$PWD" && python models/export.py --weights ./weights/yolov5s.pt --img 640 --batch 1
"""
import argparse
import sys
import time
sys.path.append('./') # to run '$ python *.py' files in subdirectories
import torch
import torch.nn as nn
from yoloface.models.experimental import attempt_load
from yoloface.models.common import Conv
from yoloface.utils.activations import Hardswish, SiLU
from yoloface.utils.general import set_logging, check_img_size
import onnx
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--weights', type=str, default='./yolov5s.pt', help='weights path') # from yolov5/models/
parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='image size') # height, width
parser.add_argument('--batch-size', type=int, default=1, help='batch size')
opt = parser.parse_args()
opt.img_size *= 2 if len(opt.img_size) == 1 else 1 # expand
print(opt)
set_logging()
t = time.time()
# Load PyTorch model
model = attempt_load(opt.weights, map_location=torch.device('cpu')) # load FP32 model
model.eval()
labels = model.names
# Checks
gs = int(max(model.stride)) # grid size (max stride)
opt.img_size = [check_img_size(x, gs) for x in opt.img_size] # verify img_size are gs-multiples
# Input
img = torch.zeros(opt.batch_size, 3, *opt.img_size) # image size(1,3,320,192) iDetection
# Update model
for k, m in model.named_modules():
m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility
if isinstance(m, Conv): # assign export-friendly activations
if isinstance(m.act, nn.Hardswish):
m.act = Hardswish()
elif isinstance(m.act, nn.SiLU):
m.act = SiLU()
# elif isinstance(m, models.yolo.Detect):
# m.forward = m.forward_export # assign forward (optional)
model.model[-1].export = True # set Detect() layer export=True
y = model(img) # dry run
# ONNX export
print('\nStarting ONNX export with onnx %s...' % onnx.__version__)
f = opt.weights.replace('.pt', '.onnx') # filename
model.fuse() # only for ONNX
torch.onnx.export(model, img, f, verbose=False, opset_version=12, input_names=['data'],
output_names=['stride_' + str(int(x)) for x in model.stride])
# Checks
onnx_model = onnx.load(f) # load onnx model
onnx.checker.check_model(onnx_model) # check onnx model
# print(onnx.helper.printable_graph(onnx_model.graph)) # print a human readable model
print('ONNX export success, saved as %s' % f)
# Finish
print('\nExport complete (%.2fs). Visualize with https://github.com/lutzroeder/netron.' % (time.time() - t))
================================================
FILE: models/yolo.py
================================================
import argparse
import logging
import math
import sys
from copy import deepcopy
from pathlib import Path
import torch
import torch.nn as nn
sys.path.append('./') # to run '$ python *.py' files in subdirectories
logger = logging.getLogger(__name__)
from models.common import Conv, Bottleneck, SPP, DWConv, Focus, BottleneckCSP, C3, ShuffleV2Block, Concat, NMS, autoShape, StemBlock
from models.experimental import MixConv2d, CrossConv
from utils.autoanchor import check_anchor_order
from utils.general import make_divisible, check_file, set_logging
from utils.torch_utils import time_synchronized, fuse_conv_and_bn, model_info, scale_img, initialize_weights, \
select_device, copy_attr
try:
import thop # for FLOPS computation
except ImportError:
thop = None
class Detect(nn.Module):
stride = None # strides computed during build
export = False # onnx export
def __init__(self, nc=80, anchors=(), ch=()): # detection layer
super(Detect, self).__init__()
self.nc = nc # number of classes
#self.no = nc + 5 # number of outputs per anchor
self.no = nc + 5 + 10 # number of outputs per anchor
self.nl = len(anchors) # number of detection layers
self.na = len(anchors[0]) // 2 # number of anchors
self.grid = [torch.zeros(1)] * self.nl # init grid
a = torch.tensor(anchors).float().view(self.nl, -1, 2)
self.register_buffer('anchors', a) # shape(nl,na,2)
self.register_buffer('anchor_grid', a.clone().view(self.nl, 1, -1, 1, 1, 2)) # shape(nl,1,na,1,1,2)
self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch) # output conv
def forward(self, x):
# x = x.copy() # for profiling
z = [] # inference output
# self.training |= self.export
if self.export:
for i in range(self.nl):
x[i] = self.m[i](x[i])
return x
for i in range(self.nl):
x[i] = self.m[i](x[i]) # conv
bs, _, ny, nx = x[i].shape # x(bs,255,20,20) to x(bs,3,20,20,85)
x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
if not self.training: # inference
if self.grid[i].shape[2:4] != x[i].shape[2:4]:
self.grid[i] = self._make_grid(nx, ny).to(x[i].device)
y = torch.full_like(x[i], 0)
y[..., [0,1,2,3,4,15]] = x[i][..., [0,1,2,3,4,15]].sigmoid()
y[..., 5:15] = x[i][..., 5:15]
#y = x[i].sigmoid()
y[..., 0:2] = (y[..., 0:2] * 2. - 0.5 + self.grid[i].to(x[i].device)) * self.stride[i] # xy
y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh
#y[..., 5:15] = y[..., 5:15] * 8 - 4
y[..., 5:7] = y[..., 5:7] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i] # landmark x1 y1
y[..., 7:9] = y[..., 7:9] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x2 y2
y[..., 9:11] = y[..., 9:11] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x3 y3
y[..., 11:13] = y[..., 11:13] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x4 y4
y[..., 13:15] = y[..., 13:15] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x5 y5
#y[..., 5:7] = (y[..., 5:7] * 2 -1) * self.anchor_grid[i] # landmark x1 y1
#y[..., 7:9] = (y[..., 7:9] * 2 -1) * self.anchor_grid[i] # landmark x2 y2
#y[..., 9:11] = (y[..., 9:11] * 2 -1) * self.anchor_grid[i] # landmark x3 y3
#y[..., 11:13] = (y[..., 11:13] * 2 -1) * self.anchor_grid[i] # landmark x4 y4
#y[..., 13:15] = (y[..., 13:15] * 2 -1) * self.anchor_grid[i] # landmark x5 y5
z.append(y.view(bs, -1, self.no))
return x if self.training else (torch.cat(z, 1), x)
@staticmethod
def _make_grid(nx=20, ny=20):
yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)], indexing='ij')
return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float()
class Model(nn.Module):
def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None): # model, input channels, number of classes
super(Model, self).__init__()
if isinstance(cfg, dict):
self.yaml = cfg # model dict
else: # is *.yaml
import yaml # for torch hub
self.yaml_file = Path(cfg).name
with open(cfg) as f:
self.yaml = yaml.load(f, Loader=yaml.FullLoader) # model dict
# Define model
ch = self.yaml['ch'] = self.yaml.get('ch', ch) # input channels
if nc and nc != self.yaml['nc']:
logger.info('Overriding model.yaml nc=%g with nc=%g' % (self.yaml['nc'], nc))
self.yaml['nc'] = nc # override yaml value
self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist
self.names = [str(i) for i in range(self.yaml['nc'])] # default names
# print([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))])
# Build strides, anchors
m = self.model[-1] # Detect()
if isinstance(m, Detect):
s = 128 # 2x min stride
m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward
m.anchors /= m.stride.view(-1, 1, 1)
check_anchor_order(m)
self.stride = m.stride
self._initialize_biases() # only run once
# print('Strides: %s' % m.stride.tolist())
# Init weights, biases
initialize_weights(self)
self.info()
logger.info('')
def forward(self, x, augment=False, profile=False):
if augment:
img_size = x.shape[-2:] # height, width
s = [1, 0.83, 0.67] # scales
f = [None, 3, None] # flips (2-ud, 3-lr)
y = [] # outputs
for si, fi in zip(s, f):
xi = scale_img(x.flip(fi) if fi else x, si)
yi = self.forward_once(xi)[0] # forward
# cv2.imwrite('img%g.jpg' % s, 255 * xi[0].numpy().transpose((1, 2, 0))[:, :, ::-1]) # save
yi[..., :4] /= si # de-scale
if fi == 2:
yi[..., 1] = img_size[0] - yi[..., 1] # de-flip ud
elif fi == 3:
yi[..., 0] = img_size[1] - yi[..., 0] # de-flip lr
y.append(yi)
return torch.cat(y, 1), None # augmented inference, train
else:
return self.forward_once(x, profile) # single-scale inference, train
def forward_once(self, x, profile=False):
y, dt = [], [] # outputs
for m in self.model:
if m.f != -1: # if not from previous layer
x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers
if profile:
o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPS
t = time_synchronized()
for _ in range(10):
_ = m(x)
dt.append((time_synchronized() - t) * 100)
print('%10.1f%10.0f%10.1fms %-40s' % (o, m.np, dt[-1], m.type))
x = m(x) # run
y.append(x if m.i in self.save else None) # save output
if profile:
print('%.1fms total' % sum(dt))
return x
def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency
# https://arxiv.org/abs/1708.02002 section 3.3
# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.
m = self.model[-1] # Detect() module
for mi, s in zip(m.m, m.stride): # from
b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85)
b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image)
b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls
mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
def _print_biases(self):
m = self.model[-1] # Detect() module
for mi in m.m: # from
b = mi.bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85)
print(('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))
# def _print_weights(self):
# for m in self.model.modules():
# if type(m) is Bottleneck:
# print('%10.3g' % (m.w.detach().sigmoid() * 2)) # shortcut weights
def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers
print('Fusing layers... ')
for m in self.model.modules():
if type(m) is Conv and hasattr(m, 'bn'):
m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv
delattr(m, 'bn') # remove batchnorm
m.forward = m.fuseforward # update forward
self.info()
return self
def nms(self, mode=True): # add or remove NMS module
present = type(self.model[-1]) is NMS # last layer is NMS
if mode and not present:
print('Adding NMS... ')
m = NMS() # module
m.f = -1 # from
m.i = self.model[-1].i + 1 # index
self.model.add_module(name='%s' % m.i, module=m) # add
self.eval()
elif not mode and present:
print('Removing NMS... ')
self.model = self.model[:-1] # remove
return self
def autoshape(self): # add autoShape module
print('Adding autoShape... ')
m = autoShape(self) # wrap model
copy_attr(m, self, include=('yaml', 'nc', 'hyp', 'names', 'stride'), exclude=()) # copy attributes
return m
def info(self, verbose=False, img_size=640): # print model information
model_info(self, verbose, img_size)
def parse_model(d, ch): # model_dict, input_channels(3)
# logger.info('\n%3s%18s%3s%10s %-40s%-30s' % ('', 'from', 'n', 'params', 'module', 'arguments'))
anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors # number of anchors
no = na * (nc + 5) # number of outputs = anchors * (classes + 5)
layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out
for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # from, number, module, args
m = eval(m) if isinstance(m, str) else m # eval strings
for j, a in enumerate(args):
try:
args[j] = eval(a) if isinstance(a, str) else a # eval strings
except:
pass
n = max(round(n * gd), 1) if n > 1 else n # depth gain
if m in [Conv, Bottleneck, SPP, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, ShuffleV2Block, StemBlock]:
c1, c2 = ch[f], args[0]
# Normal
# if i > 0 and args[0] != no: # channel expansion factor
# ex = 1.75 # exponential (default 2.0)
# e = math.log(c2 / ch[1]) / math.log(2)
# c2 = int(ch[1] * ex ** e)
# if m != Focus:
c2 = make_divisible(c2 * gw, 8) if c2 != no else c2
# Experimental
# if i > 0 and args[0] != no: # channel expansion factor
# ex = 1 + gw # exponential (default 2.0)
# ch1 = 32 # ch[1]
# e = math.log(c2 / ch1) / math.log(2) # level 1-n
# c2 = int(ch1 * ex ** e)
# if m != Focus:
# c2 = make_divisible(c2, 8) if c2 != no else c2
args = [c1, c2, *args[1:]]
if m in [BottleneckCSP, C3]:
args.insert(2, n)
n = 1
elif m is nn.BatchNorm2d:
args = [ch[f]]
elif m is Concat:
c2 = sum([ch[-1 if x == -1 else x + 1] for x in f])
elif m is Detect:
args.append([ch[x + 1] for x in f])
if isinstance(args[1], int): # number of anchors
args[1] = [list(range(args[1] * 2))] * len(f)
else:
c2 = ch[f]
m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n > 1 else m(*args) # module
t = str(m)[8:-2].replace('__main__.', '') # module type
np = sum([x.numel() for x in m_.parameters()]) # number params
m_.i, m_.f, m_.type, m_.np = i, f, t, np # attach index, 'from' index, type, number params
# logger.info('%3s%18s%3s%10.0f %-40s%-30s' % (i, f, n, np, t, args)) # print
save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelist
layers.append(m_)
ch.append(c2)
return nn.Sequential(*layers), sorted(save)
from thop import profile
from thop import clever_format
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml')
parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
opt = parser.parse_args()
opt.cfg = check_file(opt.cfg) # check file
set_logging()
device = select_device(opt.device)
# Create model
model = Model(opt.cfg).to(device)
stride = model.stride.max()
if stride == 32:
input = torch.Tensor(1, 3, 480, 640).to(device)
else:
input = torch.Tensor(1, 3, 512, 640).to(device)
model.train()
# print(model)
flops, params = profile(model, inputs=(input, ))
flops, params = clever_format([flops, params], "%.3f")
print('Flops:', flops, ',Params:' ,params)
================================================
FILE: models/yolov5-0.5.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 0.5 # layer channel multiple
# anchors
anchors:
- [4,5, 8,10, 13,16] # P3/8
- [23,29, 43,55, 73,105] # P4/16
- [146,217, 231,300, 335,433] # P5/32
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [32, 3, 2]], # 0-P2/4
[-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
[-1, 3, ShuffleV2Block, [128, 1]], # 2
[-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
[-1, 7, ShuffleV2Block, [256, 1]], # 4
[-1, 1, ShuffleV2Block, [512, 2]], # 5-P5/32
[-1, 3, ShuffleV2Block, [512, 1]], # 6
]
# YOLOv5 head
head:
[[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P4
[-1, 1, C3, [128, False]], # 10
[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 2], 1, Concat, [1]], # cat backbone P3
[-1, 1, C3, [128, False]], # 14 (P3/8-small)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 11], 1, Concat, [1]], # cat head P4
[-1, 1, C3, [128, False]], # 17 (P4/16-medium)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 7], 1, Concat, [1]], # cat head P5
[-1, 1, C3, [128, False]], # 20 (P5/32-large)
[[14, 17, 20], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5l.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 1.0 # layer channel multiple
# anchors
anchors:
- [4,5, 8,10, 13,16] # P3/8
- [23,29, 43,55, 73,105] # P4/16
- [146,217, 231,300, 335,433] # P5/32
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [64, 3, 2]], # 0-P1/2
[-1, 3, C3, [128]],
[-1, 1, Conv, [256, 3, 2]], # 2-P3/8
[-1, 9, C3, [256]],
[-1, 1, Conv, [512, 3, 2]], # 4-P4/16
[-1, 9, C3, [512]],
[-1, 1, Conv, [1024, 3, 2]], # 6-P5/32
[-1, 1, SPP, [1024, [3,5,7]]],
[-1, 3, C3, [1024, False]], # 8
]
# YOLOv5 head
head:
[[-1, 1, Conv, [512, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 5], 1, Concat, [1]], # cat backbone P4
[-1, 3, C3, [512, False]], # 12
[-1, 1, Conv, [256, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 3], 1, Concat, [1]], # cat backbone P3
[-1, 3, C3, [256, False]], # 16 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]],
[[-1, 13], 1, Concat, [1]], # cat head P4
[-1, 3, C3, [512, False]], # 19 (P4/16-medium)
[-1, 1, Conv, [512, 3, 2]],
[[-1, 9], 1, Concat, [1]], # cat head P5
[-1, 3, C3, [1024, False]], # 22 (P5/32-large)
[[16, 19, 22], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5l6.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 1.0 # layer channel multiple
# anchors
anchors:
- [6,7, 9,11, 13,16] # P3/8
- [18,23, 26,33, 37,47] # P4/16
- [54,67, 77,104, 112,154] # P5/32
- [174,238, 258,355, 445,568] # P6/64
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[ [ -1, 1, StemBlock, [ 64, 3, 2 ] ], # 0-P1/2
[ -1, 3, C3, [ 128 ] ],
[ -1, 1, Conv, [ 256, 3, 2 ] ], # 2-P3/8
[ -1, 9, C3, [ 256 ] ],
[ -1, 1, Conv, [ 512, 3, 2 ] ], # 4-P4/16
[ -1, 9, C3, [ 512 ] ],
[ -1, 1, Conv, [ 768, 3, 2 ] ], # 6-P5/32
[ -1, 3, C3, [ 768 ] ],
[ -1, 1, Conv, [ 1024, 3, 2 ] ], # 8-P6/64
[ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
[ -1, 3, C3, [ 1024, False ] ], # 10
]
# YOLOv5 head
head:
[ [ -1, 1, Conv, [ 768, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 7 ], 1, Concat, [ 1 ] ], # cat backbone P5
[ -1, 3, C3, [ 768, False ] ], # 14
[ -1, 1, Conv, [ 512, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 5 ], 1, Concat, [ 1 ] ], # cat backbone P4
[ -1, 3, C3, [ 512, False ] ], # 18
[ -1, 1, Conv, [ 256, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 3 ], 1, Concat, [ 1 ] ], # cat backbone P3
[ -1, 3, C3, [ 256, False ] ], # 22 (P3/8-small)
[ -1, 1, Conv, [ 256, 3, 2 ] ],
[ [ -1, 19 ], 1, Concat, [ 1 ] ], # cat head P4
[ -1, 3, C3, [ 512, False ] ], # 25 (P4/16-medium)
[ -1, 1, Conv, [ 512, 3, 2 ] ],
[ [ -1, 15 ], 1, Concat, [ 1 ] ], # cat head P5
[ -1, 3, C3, [ 768, False ] ], # 28 (P5/32-large)
[ -1, 1, Conv, [ 768, 3, 2 ] ],
[ [ -1, 11 ], 1, Concat, [ 1 ] ], # cat head P6
[ -1, 3, C3, [ 1024, False ] ], # 31 (P6/64-xlarge)
[ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6)
]
================================================
FILE: models/yolov5m.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 0.67 # model depth multiple
width_multiple: 0.75 # layer channel multiple
# anchors
anchors:
- [4,5, 8,10, 13,16] # P3/8
- [23,29, 43,55, 73,105] # P4/16
- [146,217, 231,300, 335,433] # P5/32
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [64, 3, 2]], # 0-P1/2
[-1, 3, C3, [128]],
[-1, 1, Conv, [256, 3, 2]], # 2-P3/8
[-1, 9, C3, [256]],
[-1, 1, Conv, [512, 3, 2]], # 4-P4/16
[-1, 9, C3, [512]],
[-1, 1, Conv, [1024, 3, 2]], # 6-P5/32
[-1, 1, SPP, [1024, [3,5,7]]],
[-1, 3, C3, [1024, False]], # 8
]
# YOLOv5 head
head:
[[-1, 1, Conv, [512, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 5], 1, Concat, [1]], # cat backbone P4
[-1, 3, C3, [512, False]], # 12
[-1, 1, Conv, [256, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 3], 1, Concat, [1]], # cat backbone P3
[-1, 3, C3, [256, False]], # 16 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]],
[[-1, 13], 1, Concat, [1]], # cat head P4
[-1, 3, C3, [512, False]], # 19 (P4/16-medium)
[-1, 1, Conv, [512, 3, 2]],
[[-1, 9], 1, Concat, [1]], # cat head P5
[-1, 3, C3, [1024, False]], # 22 (P5/32-large)
[[16, 19, 22], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5m6.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 0.67 # model depth multiple
width_multiple: 0.75 # layer channel multiple
# anchors
anchors:
- [6,7, 9,11, 13,16] # P3/8
- [18,23, 26,33, 37,47] # P4/16
- [54,67, 77,104, 112,154] # P5/32
- [174,238, 258,355, 445,568] # P6/64
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[ [ -1, 1, StemBlock, [ 64, 3, 2 ] ], # 0-P1/2
[ -1, 3, C3, [ 128 ] ],
[ -1, 1, Conv, [ 256, 3, 2 ] ], # 2-P3/8
[ -1, 9, C3, [ 256 ] ],
[ -1, 1, Conv, [ 512, 3, 2 ] ], # 4-P4/16
[ -1, 9, C3, [ 512 ] ],
[ -1, 1, Conv, [ 768, 3, 2 ] ], # 6-P5/32
[ -1, 3, C3, [ 768 ] ],
[ -1, 1, Conv, [ 1024, 3, 2 ] ], # 8-P6/64
[ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
[ -1, 3, C3, [ 1024, False ] ], # 10
]
# YOLOv5 head
head:
[ [ -1, 1, Conv, [ 768, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 7 ], 1, Concat, [ 1 ] ], # cat backbone P5
[ -1, 3, C3, [ 768, False ] ], # 14
[ -1, 1, Conv, [ 512, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 5 ], 1, Concat, [ 1 ] ], # cat backbone P4
[ -1, 3, C3, [ 512, False ] ], # 18
[ -1, 1, Conv, [ 256, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 3 ], 1, Concat, [ 1 ] ], # cat backbone P3
[ -1, 3, C3, [ 256, False ] ], # 22 (P3/8-small)
[ -1, 1, Conv, [ 256, 3, 2 ] ],
[ [ -1, 19 ], 1, Concat, [ 1 ] ], # cat head P4
[ -1, 3, C3, [ 512, False ] ], # 25 (P4/16-medium)
[ -1, 1, Conv, [ 512, 3, 2 ] ],
[ [ -1, 15 ], 1, Concat, [ 1 ] ], # cat head P5
[ -1, 3, C3, [ 768, False ] ], # 28 (P5/32-large)
[ -1, 1, Conv, [ 768, 3, 2 ] ],
[ [ -1, 11 ], 1, Concat, [ 1 ] ], # cat head P6
[ -1, 3, C3, [ 1024, False ] ], # 31 (P6/64-xlarge)
[ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6)
]
================================================
FILE: models/yolov5n.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 1.0 # layer channel multiple
# anchors
anchors:
- [4,5, 8,10, 13,16] # P3/8
- [23,29, 43,55, 73,105] # P4/16
- [146,217, 231,300, 335,433] # P5/32
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [32, 3, 2]], # 0-P2/4
[-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
[-1, 3, ShuffleV2Block, [128, 1]], # 2
[-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
[-1, 7, ShuffleV2Block, [256, 1]], # 4
[-1, 1, ShuffleV2Block, [512, 2]], # 5-P5/32
[-1, 3, ShuffleV2Block, [512, 1]], # 6
]
# YOLOv5 head
head:
[[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P4
[-1, 1, C3, [128, False]], # 10
[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 2], 1, Concat, [1]], # cat backbone P3
[-1, 1, C3, [128, False]], # 14 (P3/8-small)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 11], 1, Concat, [1]], # cat head P4
[-1, 1, C3, [128, False]], # 17 (P4/16-medium)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 7], 1, Concat, [1]], # cat head P5
[-1, 1, C3, [128, False]], # 20 (P5/32-large)
[[14, 17, 20], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5n6.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 1.0 # model depth multiple
width_multiple: 1.0 # layer channel multiple
# anchors
anchors:
- [6,7, 9,11, 13,16] # P3/8
- [18,23, 26,33, 37,47] # P4/16
- [54,67, 77,104, 112,154] # P5/32
- [174,238, 258,355, 445,568] # P6/64
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [32, 3, 2]], # 0-P2/4
[-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
[-1, 3, ShuffleV2Block, [128, 1]], # 2
[-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
[-1, 7, ShuffleV2Block, [256, 1]], # 4
[-1, 1, ShuffleV2Block, [384, 2]], # 5-P5/32
[-1, 3, ShuffleV2Block, [384, 1]], # 6
[-1, 1, ShuffleV2Block, [512, 2]], # 7-P6/64
[-1, 3, ShuffleV2Block, [512, 1]], # 8
]
# YOLOv5 head
head:
[[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 6], 1, Concat, [1]], # cat backbone P5
[-1, 1, C3, [128, False]], # 12
[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 4], 1, Concat, [1]], # cat backbone P4
[-1, 1, C3, [128, False]], # 16 (P4/8-small)
[-1, 1, Conv, [128, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 2], 1, Concat, [1]], # cat backbone P3
[-1, 1, C3, [128, False]], # 20 (P3/8-small)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 17], 1, Concat, [1]], # cat head P4
[-1, 1, C3, [128, False]], # 23 (P4/16-medium)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 13], 1, Concat, [1]], # cat head P5
[-1, 1, C3, [128, False]], # 26 (P5/32-large)
[-1, 1, Conv, [128, 3, 2]],
[[-1, 9], 1, Concat, [1]], # cat head P6
[-1, 1, C3, [128, False]], # 29 (P6/64-large)
[[20, 23, 26, 29], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5s.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.35 # layer channel multiple
# anchors
anchors:
- [4,5, 8,10, 13,16] # P3/8
- [23,29, 43,55, 73,105] # P4/16
- [146,217, 231,300, 335,433] # P5/32
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[[-1, 1, StemBlock, [64, 3, 2]], # 0-P1/2
[-1, 3, C3, [128]],
[-1, 1, Conv, [256, 3, 2]], # 2-P3/8
[-1, 9, C3, [256]],
[-1, 1, Conv, [512, 3, 2]], # 4-P4/16
[-1, 9, C3, [512]],
[-1, 1, Conv, [1024, 3, 2]], # 6-P5/32
[-1, 1, SPP, [1024, [3,5,7]]],
[-1, 3, C3, [1024, False]], # 8
]
# YOLOv5 head
head:
[[-1, 1, Conv, [512, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 5], 1, Concat, [1]], # cat backbone P4
[-1, 3, C3, [512, False]], # 12
[-1, 1, Conv, [256, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']],
[[-1, 3], 1, Concat, [1]], # cat backbone P3
[-1, 3, C3, [256, False]], # 16 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]],
[[-1, 13], 1, Concat, [1]], # cat head P4
[-1, 3, C3, [512, False]], # 19 (P4/16-medium)
[-1, 1, Conv, [512, 3, 2]],
[[-1, 9], 1, Concat, [1]], # cat head P5
[-1, 3, C3, [1024, False]], # 22 (P5/32-large)
[[16, 19, 22], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
================================================
FILE: models/yolov5s6.yaml
================================================
# parameters
nc: 1 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.50 # layer channel multiple
# anchors
anchors:
- [6,7, 9,11, 13,16] # P3/8
- [18,23, 26,33, 37,47] # P4/16
- [54,67, 77,104, 112,154] # P5/32
- [174,238, 258,355, 445,568] # P6/64
# YOLOv5 backbone
backbone:
# [from, number, module, args]
[ [ -1, 1, StemBlock, [ 64, 3, 2 ] ], # 0-P1/2
[ -1, 3, C3, [ 128 ] ],
[ -1, 1, Conv, [ 256, 3, 2 ] ], # 2-P3/8
[ -1, 9, C3, [ 256 ] ],
[ -1, 1, Conv, [ 512, 3, 2 ] ], # 4-P4/16
[ -1, 9, C3, [ 512 ] ],
[ -1, 1, Conv, [ 768, 3, 2 ] ], # 6-P5/32
[ -1, 3, C3, [ 768 ] ],
[ -1, 1, Conv, [ 1024, 3, 2 ] ], # 8-P6/64
[ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
[ -1, 3, C3, [ 1024, False ] ], # 10
]
# YOLOv5 head
head:
[ [ -1, 1, Conv, [ 768, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 7 ], 1, Concat, [ 1 ] ], # cat backbone P5
[ -1, 3, C3, [ 768, False ] ], # 14
[ -1, 1, Conv, [ 512, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 5 ], 1, Concat, [ 1 ] ], # cat backbone P4
[ -1, 3, C3, [ 512, False ] ], # 18
[ -1, 1, Conv, [ 256, 1, 1 ] ],
[ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
[ [ -1, 3 ], 1, Concat, [ 1 ] ], # cat backbone P3
[ -1, 3, C3, [ 256, False ] ], # 22 (P3/8-small)
[ -1, 1, Conv, [ 256, 3, 2 ] ],
[ [ -1, 19 ], 1, Concat, [ 1 ] ], # cat head P4
[ -1, 3, C3, [ 512, False ] ], # 25 (P4/16-medium)
[ -1, 1, Conv, [ 512, 3, 2 ] ],
[ [ -1, 15 ], 1, Concat, [ 1 ] ], # cat head P5
[ -1, 3, C3, [ 768, False ] ], # 28 (P5/32-large)
[ -1, 1, Conv, [ 768, 3, 2 ] ],
[ [ -1, 11 ], 1, Concat, [ 1 ] ], # cat head P6
[ -1, 3, C3, [ 1024, False ] ], # 31 (P6/64-xlarge)
[ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6)
]
================================================
FILE: requirements.txt
================================================
joblib==1.2.0
matplotlib==3.5.1
numpy==1.22.4
onnx==1.12.0
opencv_python==4.6.0.66
pandas==1.4.2
Pillow==9.3.0
PyYAML==6.0
requests==2.27.1
scipy==1.7.3
seaborn==0.11.2
setuptools==61.2.0
thop==0.1.1.post2207130030
torch==1.12.1+cu116
torchvision==0.13.1+cu116
tqdm==4.64.0
wandb==0.13.6
================================================
FILE: utils/__init__.py
================================================
================================================
FILE: utils/activations.py
================================================
# Activation functions
import torch
import torch.nn as nn
import torch.nn.functional as F
# SiLU https://arxiv.org/pdf/1606.08415.pdf ----------------------------------------------------------------------------
class SiLU(nn.Module): # export-friendly version of nn.SiLU()
@staticmethod
def forward(x):
return x * torch.sigmoid(x)
class Hardswish(nn.Module): # export-friendly version of nn.Hardswish()
@staticmethod
def forward(x):
# return x * F.hardsigmoid(x) # for torchscript and CoreML
return x * F.hardtanh(x + 3, 0., 6.) / 6. # for torchscript, CoreML and ONNX
class MemoryEfficientSwish(nn.Module):
class F(torch.autograd.Function):
@staticmethod
def forward(ctx, x):
ctx.save_for_backward(x)
return x * torch.sigmoid(x)
@staticmethod
def backward(ctx, grad_output):
x = ctx.saved_tensors[0]
sx = torch.sigmoid(x)
return grad_output * (sx * (1 + x * (1 - sx)))
def forward(self, x):
return self.F.apply(x)
# Mish https://github.com/digantamisra98/Mish --------------------------------------------------------------------------
class Mish(nn.Module):
@staticmethod
def forward(x):
return x * F.softplus(x).tanh()
class MemoryEfficientMish(nn.Module):
class F(torch.autograd.Function):
@staticmethod
def forward(ctx, x):
ctx.save_for_backward(x)
return x.mul(torch.tanh(F.softplus(x))) # x * tanh(ln(1 + exp(x)))
@staticmethod
def backward(ctx, grad_output):
x = ctx.saved_tensors[0]
sx = torch.sigmoid(x)
fx = F.softplus(x).tanh()
return grad_output * (fx + x * sx * (1 - fx * fx))
def forward(self, x):
return self.F.apply(x)
# FReLU https://arxiv.org/abs/2007.11824 -------------------------------------------------------------------------------
class FReLU(nn.Module):
def __init__(self, c1, k=3): # ch_in, kernel
super().__init__()
self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False)
self.bn = nn.BatchNorm2d(c1)
def forward(self, x):
return torch.max(x, self.bn(self.conv(x)))
================================================
FILE: utils/autoanchor.py
================================================
# Auto-anchor utils
import numpy as np
import torch
import yaml
from scipy.cluster.vq import kmeans
from tqdm import tqdm
from utils.general import colorstr
def check_anchor_order(m):
# Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary
a = m.anchor_grid.prod(-1).view(-1) # anchor area
da = a[-1] - a[0] # delta a
ds = m.stride[-1] - m.stride[0] # delta s
if da.sign() != ds.sign(): # same order
print('Reversing anchor order')
m.anchors[:] = m.anchors.flip(0)
m.anchor_grid[:] = m.anchor_grid.flip(0)
def check_anchors(dataset, model, thr=4.0, imgsz=640):
# Check anchor fit to data, recompute if necessary
prefix = colorstr('autoanchor: ')
print(f'\n{prefix}Analyzing anchors... ', end='')
m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect()
shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True)
scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1)) # augment scale
wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float() # wh
def metric(k): # compute metric
r = wh[:, None] / k[None]
x = torch.min(r, 1. / r).min(2)[0] # ratio metric
best = x.max(1)[0] # best_x
aat = (x > 1. / thr).float().sum(1).mean() # anchors above threshold
bpr = (best > 1. / thr).float().mean() # best possible recall
return bpr, aat
bpr, aat = metric(m.anchor_grid.clone().cpu().view(-1, 2))
print(f'anchors/target = {aat:.2f}, Best Possible Recall (BPR) = {bpr:.4f}', end='')
if bpr < 0.98: # threshold to recompute
print('. Attempting to improve anchors, please wait...')
na = m.anchor_grid.numel() // 2 # number of anchors
new_anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False)
new_bpr = metric(new_anchors.reshape(-1, 2))[0]
if new_bpr > bpr: # replace anchors
new_anchors = torch.tensor(new_anchors, device=m.anchors.device).type_as(m.anchors)
m.anchor_grid[:] = new_anchors.clone().view_as(m.anchor_grid) # for inference
m.anchors[:] = new_anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss
check_anchor_order(m)
print(f'{prefix}New anchors saved to model. Update model *.yaml to use these anchors in the future.')
else:
print(f'{prefix}Original anchors better than new anchors. Proceeding with original anchors.')
print('') # newline
def kmean_anchors(path='./data/coco128.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True):
""" Creates kmeans-evolved anchors from training dataset
Arguments:
path: path to dataset *.yaml, or a loaded dataset
n: number of anchors
img_size: image size used for training
thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0
gen: generations to evolve anchors using genetic algorithm
verbose: print all results
Return:
k: kmeans evolved anchors
Usage:
from utils.autoanchor import *; _ = kmean_anchors()
"""
thr = 1. / thr
prefix = colorstr('autoanchor: ')
def metric(k, wh): # compute metrics
r = wh[:, None] / k[None]
x = torch.min(r, 1. / r).min(2)[0] # ratio metric
# x = wh_iou(wh, torch.tensor(k)) # iou metric
return x, x.max(1)[0] # x, best_x
def anchor_fitness(k): # mutation fitness
_, best = metric(torch.tensor(k, dtype=torch.float32), wh)
return (best * (best > thr).float()).mean() # fitness
def print_results(k):
k = k[np.argsort(k.prod(1))] # sort small to large
x, best = metric(k, wh0)
bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n # best possible recall, anch > thr
print(f'{prefix}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr')
print(f'{prefix}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, '
f'past_thr={x[x > thr].mean():.3f}-mean: ', end='')
for i, x in enumerate(k):
print('%i,%i' % (round(x[0]), round(x[1])), end=', ' if i < len(k) - 1 else '\n') # use in *.cfg
return k
if isinstance(path, str): # *.yaml file
with open(path) as f:
data_dict = yaml.load(f, Loader=yaml.SafeLoader) # model dict
from utils.datasets import LoadImagesAndLabels
dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True)
else:
dataset = path # dataset
# Get label wh
shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True)
wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)]) # wh
# Filter
i = (wh0 < 3.0).any(1).sum()
if i:
print(f'{prefix}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.')
wh = wh0[(wh0 >= 2.0).any(1)] # filter > 2 pixels
# wh = wh * (np.random.rand(wh.shape[0], 1) * 0.9 + 0.1) # multiply by random scale 0-1
# Kmeans calculation
print(f'{prefix}Running kmeans for {n} anchors on {len(wh)} points...')
s = wh.std(0) # sigmas for whitening
k, dist = kmeans(wh / s, n, iter=30) # points, mean distance
k *= s
wh = torch.tensor(wh, dtype=torch.float32) # filtered
wh0 = torch.tensor(wh0, dtype=torch.float32) # unfiltered
k = print_results(k)
# Plot
# k, d = [None] * 20, [None] * 20
# for i in tqdm(range(1, 21)):
# k[i-1], d[i-1] = kmeans(wh / s, i) # points, mean distance
# fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True)
# ax = ax.ravel()
# ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.')
# fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # plot wh
# ax[0].hist(wh[wh[:, 0]<100, 0],400)
# ax[1].hist(wh[wh[:, 1]<100, 1],400)
# fig.savefig('wh.png', dpi=200)
# Evolve
npr = np.random
f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1 # fitness, generations, mutation prob, sigma
pbar = tqdm(range(gen), desc=f'{prefix}Evolving anchors with Genetic Algorithm:') # progress bar
for _ in pbar:
v = np.ones(sh)
while (v == 1).all(): # mutate until a change occurs (prevent duplicates)
v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0)
kg = (k.copy() * v).clip(min=2.0)
fg = anchor_fitness(kg)
if fg > f:
f, k = fg, kg.copy()
pbar.desc = f'{prefix}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}'
if verbose:
print_results(k)
return print_results(k)
================================================
FILE: utils/aws/__init__.py
================================================
================================================
FILE: utils/aws/mime.sh
================================================
# AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/
# This script will run on every instance restart, not only on first start
# --- DO NOT COPY ABOVE COMMENTS WHEN PASTING INTO USERDATA ---
Content-Type: multipart/mixed; boundary="//"
MIME-Version: 1.0
--//
Content-Type: text/cloud-config; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename="cloud-config.txt"
#cloud-config
cloud_final_modules:
- [scripts-user, always]
--//
Content-Type: text/x-shellscript; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename="userdata.txt"
#!/bin/bash
# --- paste contents of userdata.sh here ---
--//
================================================
FILE: utils/aws/resume.py
================================================
# Resume all interrupted trainings in yolov5/ dir including DDP trainings
# Usage: $ python utils/aws/resume.py
import os
import sys
from pathlib import Path
import torch
import yaml
sys.path.append('./') # to run '$ python *.py' files in subdirectories
port = 0 # --master_port
path = Path('').resolve()
for last in path.rglob('*/**/last.pt'):
ckpt = torch.load(last)
if ckpt['optimizer'] is None:
continue
# Load opt.yaml
with open(last.parent.parent / 'opt.yaml') as f:
opt = yaml.load(f, Loader=yaml.SafeLoader)
# Get device count
d = opt['device'].split(',') # devices
nd = len(d) # number of devices
ddp = nd > 1 or (nd == 0 and torch.cuda.device_count() > 1) # distributed data parallel
if ddp: # multi-GPU
port += 1
cmd = f'python -m torch.distributed.launch --nproc_per_node {nd} --master_port {port} train.py --resume {last}'
else: # single-GPU
cmd = f'python train.py --resume {last}'
cmd += ' > /dev/null 2>&1 &' # redirect output to dev/null and run in daemon thread
print(cmd)
os.system(cmd)
================================================
FILE: utils/aws/userdata.sh
================================================
#!/bin/bash
# AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html
# This script will run only once on first instance start (for a re-start script see mime.sh)
# /home/ubuntu (ubuntu) or /home/ec2-user (amazon-linux) is working dir
# Use >300 GB SSD
cd home/ubuntu
if [ ! -d yolov5 ]; then
echo "Running first-time script." # install dependencies, download COCO, pull Docker
git clone https://github.com/ultralytics/yolov5 && sudo chmod -R 777 yolov5
cd yolov5
bash data/scripts/get_coco.sh && echo "Data done." &
sudo docker pull ultralytics/yolov5:latest && echo "Docker done." &
python -m pip install --upgrade pip && pip install -r requirements.txt && python detect.py && echo "Requirements done." &
wait && echo "All tasks done." # finish background tasks
else
echo "Running re-start script." # resume interrupted runs
i=0
list=$(sudo docker ps -qa) # container list i.e. $'one\ntwo\nthree\nfour'
while IFS= read -r id; do
((i++))
echo "restarting container $i: $id"
sudo docker start $id
# sudo docker exec -it $id python train.py --resume # single-GPU
sudo docker exec -d $id python utils/aws/resume.py # multi-scenario
done <<<"$list"
fi
================================================
FILE: utils/datasets.py
================================================
# Dataset utils and dataloaders
import glob
import logging
import math
import os
import random
import shutil
import time
from itertools import repeat
from multiprocessing.pool import ThreadPool
from pathlib import Path
from threading import Thread
import cv2
import numpy as np
import torch
import torch.nn.functional as F
from PIL import Image, ExifTags
from torch.utils.data import Dataset
from tqdm import tqdm
from utils.general import xyxy2xywh, xywh2xyxy, xywhn2xyxy, clean_str
from utils.torch_utils import torch_distributed_zero_first
# Parameters
help_url = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
img_formats = ['bmp', 'jpg', 'jpeg', 'png', 'tif', 'tiff', 'dng'] # acceptable image suffixes
vid_formats = ['mov', 'avi', 'mp4', 'mpg', 'mpeg', 'm4v', 'wmv', 'mkv'] # acceptable video suffixes
logger = logging.getLogger(__name__)
# Get orientation exif tag
for orientation in ExifTags.TAGS.keys():
if ExifTags.TAGS[orientation] == 'Orientation':
break
def get_hash(files):
# Returns a single hash value of a list of files
return sum(os.path.getsize(f) for f in files if os.path.isfile(f))
def exif_size(img):
# Returns exif-corrected PIL size
s = img.size # (width, height)
try:
rotation = dict(img._getexif().items())[orientation]
if rotation == 6: # rotation 270
s = (s[1], s[0])
elif rotation == 8: # rotation 90
s = (s[1], s[0])
except:
pass
return s
def create_dataloader(path, imgsz, batch_size, stride, opt, hyp=None, augment=False, cache=False, pad=0.0, rect=False,
rank=-1, world_size=1, workers=8, image_weights=False, quad=False, prefix=''):
# Make sure only the first process in DDP process the dataset first, and the following others can use the cache
with torch_distributed_zero_first(rank):
dataset = LoadImagesAndLabels(path, imgsz, batch_size,
augment=augment, # augment images
hyp=hyp, # augmentation hyperparameters
rect=rect, # rectangular training
cache_images=cache,
single_cls=opt.single_cls,
stride=int(stride),
pad=pad,
image_weights=image_weights,
prefix=prefix)
batch_size = min(batch_size, len(dataset))
nw = min([os.cpu_count() // world_size, batch_size if batch_size > 1 else 0, workers]) # number of workers
sampler = torch.utils.data.distributed.DistributedSampler(dataset) if rank != -1 else None
loader = torch.utils.data.DataLoader if image_weights else InfiniteDataLoader
# Use torch.utils.data.DataLoader() if dataset.properties will update during training else InfiniteDataLoader()
dataloader = loader(dataset,
batch_size=batch_size,
num_workers=nw,
sampler=sampler,
pin_memory=True,
collate_fn=LoadImagesAndLabels.collate_fn4 if quad else LoadImagesAndLabels.collate_fn)
return dataloader, dataset
class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):
""" Dataloader that reuses workers
Uses same syntax as vanilla DataLoader
"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
object.__setattr__(self, 'batch_sampler', _RepeatSampler(self.batch_sampler))
self.iterator = super().__iter__()
def __len__(self):
return len(self.batch_sampler.sampler)
def __iter__(self):
for i in range(len(self)):
yield next(self.iterator)
class _RepeatSampler(object):
""" Sampler that repeats forever
Args:
sampler (Sampler)
"""
def __init__(self, sampler):
self.sampler = sampler
def __iter__(self):
while True:
yield from iter(self.sampler)
class LoadImages: # for inference
def __init__(self, path, img_size=640):
p = str(Path(path)) # os-agnostic
p = os.path.abspath(p) # absolute path
if '*' in p:
files = sorted(glob.glob(p, recursive=True)) # glob
elif os.path.isdir(p):
files = sorted(glob.glob(os.path.join(p, '*.*'))) # dir
elif os.path.isfile(p):
files = [p] # files
else:
raise Exception(f'ERROR: {p} does not exist')
images = [x for x in files if x.split('.')[-1].lower() in img_formats]
videos = [x for x in files if x.split('.')[-1].lower() in vid_formats]
ni, nv = len(images), len(videos)
self.img_size = img_size
self.files = images + videos
self.nf = ni + nv # number of files
self.video_flag = [False] * ni + [True] * nv
self.mode = 'image'
if any(videos):
self.new_video(videos[0]) # new video
else:
self.cap = None
assert self.nf > 0, f'No images or videos found in {p}. ' \
f'Supported formats are:\nimages: {img_formats}\nvideos: {vid_formats}'
def __iter__(self):
self.count = 0
return self
def __next__(self):
if self.count == self.nf:
raise StopIteration
path = self.files[self.count]
if self.video_flag[self.count]:
# Read video
self.mode = 'video'
ret_val, img0 = self.cap.read()
if not ret_val:
self.count += 1
self.cap.release()
if self.count == self.nf: # last video
raise StopIteration
else:
path = self.files[self.count]
self.new_video(path)
ret_val, img0 = self.cap.read()
self.frame += 1
print(f'video {self.count + 1}/{self.nf} ({self.frame}/{self.nframes}) {path}: ', end='')
else:
# Read image
self.count += 1
img0 = cv2.imread(path) # BGR
assert img0 is not None, 'Image Not Found ' + path
print(f'image {self.count}/{self.nf} {path}: ', end='')
# Padded resize
img = letterbox(img0, new_shape=self.img_size)[0]
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
return path, img, img0, self.cap
def new_video(self, path):
self.frame = 0
self.cap = cv2.VideoCapture(path)
self.nframes = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))
def __len__(self):
return self.nf # number of files
class LoadWebcam: # for inference
def __init__(self, pipe='0', img_size=640):
self.img_size = img_size
if pipe.isnumeric():
pipe = eval(pipe) # local camera
# pipe = 'rtsp://192.168.1.64/1' # IP camera
# pipe = 'rtsp://username:password@192.168.1.64/1' # IP camera with login
# pipe = 'http://wmccpinetop.axiscam.net/mjpg/video.mjpg' # IP golf camera
self.pipe = pipe
self.cap = cv2.VideoCapture(pipe) # video capture object
self.cap.set(cv2.CAP_PROP_BUFFERSIZE, 3) # set buffer size
def __iter__(self):
self.count = -1
return self
def __next__(self):
self.count += 1
if cv2.waitKey(1) == ord('q'): # q to quit
self.cap.release()
cv2.destroyAllWindows()
raise StopIteration
# Read frame
if self.pipe == 0: # local camera
ret_val, img0 = self.cap.read()
img0 = cv2.flip(img0, 1) # flip left-right
else: # IP camera
n = 0
while True:
n += 1
self.cap.grab()
if n % 30 == 0: # skip frames
ret_val, img0 = self.cap.retrieve()
if ret_val:
break
# Print
assert ret_val, f'Camera Error {self.pipe}'
img_path = 'webcam.jpg'
print(f'webcam {self.count}: ', end='')
# Padded resize
img = letterbox(img0, new_shape=self.img_size)[0]
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
return img_path, img, img0, None
def __len__(self):
return 0
class LoadStreams: # multiple IP or RTSP cameras
def __init__(self, sources='streams.txt', img_size=640):
self.mode = 'stream'
self.img_size = img_size
if os.path.isfile(sources):
with open(sources, 'r') as f:
sources = [x.strip() for x in f.read().strip().splitlines() if len(x.strip())]
else:
sources = [sources]
n = len(sources)
self.imgs = [None] * n
self.sources = [clean_str(x) for x in sources] # clean source names for later
for i, s in enumerate(sources):
# Start the thread to read frames from the video stream
print(f'{i + 1}/{n}: {s}... ', end='')
cap = cv2.VideoCapture(eval(s) if s.isnumeric() else s)
assert cap.isOpened(), f'Failed to open {s}'
w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = cap.get(cv2.CAP_PROP_FPS) % 100
_, self.imgs[i] = cap.read() # guarantee first frame
thread = Thread(target=self.update, args=([i, cap]), daemon=True)
print(f' success ({w}x{h} at {fps:.2f} FPS).')
thread.start()
print('') # newline
# check for common shapes
s = np.stack([letterbox(x, new_shape=self.img_size)[0].shape for x in self.imgs], 0) # inference shapes
self.rect = np.unique(s, axis=0).shape[0] == 1 # rect inference if all shapes equal
if not self.rect:
print('WARNING: Different stream shapes detected. For optimal performance supply similarly-shaped streams.')
def update(self, index, cap):
# Read next stream frame in a daemon thread
n = 0
while cap.isOpened():
n += 1
# _, self.imgs[index] = cap.read()
cap.grab()
if n == 4: # read every 4th frame
_, self.imgs[index] = cap.retrieve()
n = 0
time.sleep(0.01) # wait time
def __iter__(self):
self.count = -1
return self
def __next__(self):
self.count += 1
img0 = self.imgs.copy()
if cv2.waitKey(1) == ord('q'): # q to quit
cv2.destroyAllWindows()
raise StopIteration
# Letterbox
img = [letterbox(x, new_shape=self.img_size, auto=self.rect)[0] for x in img0]
# Stack
img = np.stack(img, 0)
# Convert
img = img[:, :, :, ::-1].transpose(0, 3, 1, 2) # BGR to RGB, to bsx3x416x416
img = np.ascontiguousarray(img)
return self.sources, img, img0, None
def __len__(self):
return 0 # 1E12 frames = 32 streams at 30 FPS for 30 years
def img2label_paths(img_paths):
# Define label paths as a function of image paths
sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep # /images/, /labels/ substrings
return [x.replace(sa, sb, 1).replace('.' + x.split('.')[-1], '.txt') for x in img_paths]
class LoadImagesAndLabels(Dataset): # for training/testing
def __init__(self, path, img_size=640, batch_size=16, augment=False, hyp=None, rect=False, image_weights=False,
cache_images=False, single_cls=False, stride=32, pad=0.0, prefix=''):
self.img_size = img_size
self.augment = augment
self.hyp = hyp
self.image_weights = image_weights
self.rect = False if image_weights else rect
self.mosaic = self.augment and not self.rect # load 4 images at a time into a mosaic (only during training)
self.mosaic_border = [-img_size // 2, -img_size // 2]
self.stride = stride
try:
f = [] # image files
for p in path if isinstance(path, list) else [path]:
p = Path(p) # os-agnostic
if p.is_dir(): # dir
f += glob.glob(str(p / '**' / '*.*'), recursive=True)
elif p.is_file(): # file
with open(p, 'r') as t:
t = t.read().strip().splitlines()
parent = str(p.parent) + os.sep
f += [x.replace('./', parent) if x.startswith('./') else x for x in t] # local to global path
else:
raise Exception(f'{prefix}{p} does not exist')
self.img_files = sorted([x.replace('/', os.sep) for x in f if x.split('.')[-1].lower() in img_formats])
assert self.img_files, f'{prefix}No images found'
except Exception as e:
raise Exception(f'{prefix}Error loading data from {path}: {e}\nSee {help_url}')
# Check cache
self.label_files = img2label_paths(self.img_files) # labels
cache_path = Path(self.label_files[0]).parent.with_suffix('.cache') # cached labels
if cache_path.is_file():
cache = torch.load(cache_path) # load
if cache['hash'] != get_hash(self.label_files + self.img_files) or 'results' not in cache: # changed
cache = self.cache_labels(cache_path, prefix) # re-cache
else:
cache = self.cache_labels(cache_path, prefix) # cache
# Display cache
[nf, nm, ne, nc, n] = cache.pop('results') # found, missing, empty, corrupted, total
desc = f"Scanning '{cache_path}' for images and labels... {nf} found, {nm} missing, {ne} empty, {nc} corrupted"
tqdm(None, desc=prefix + desc, total=n, initial=n)
assert nf > 0 or not augment, f'{prefix}No labels in {cache_path}. Can not train without labels. See {help_url}'
# Read cache
cache.pop('hash') # remove hash
labels, shapes = zip(*cache.values())
self.labels = list(labels)
self.shapes = np.array(shapes, dtype=np.float64)
self.img_files = list(cache.keys()) # update
self.label_files = img2label_paths(cache.keys()) # update
if single_cls:
for x in self.labels:
x[:, 0] = 0
n = len(shapes) # number of images
bi = np.floor(np.arange(n) / batch_size).astype(np.int) # batch index
nb = bi[-1] + 1 # number of batches
self.batch = bi # batch index of image
self.n = n
self.indices = range(n)
# Rectangular Training
if self.rect:
# Sort by aspect ratio
s = self.shapes # wh
ar = s[:, 1] / s[:, 0] # aspect ratio
irect = ar.argsort()
self.img_files = [self.img_files[i] for i in irect]
self.label_files = [self.label_files[i] for i in irect]
self.labels = [self.labels[i] for i in irect]
self.shapes = s[irect] # wh
ar = ar[irect]
# Set training image shapes
shapes = [[1, 1]] * nb
for i in range(nb):
ari = ar[bi == i]
mini, maxi = ari.min(), ari.max()
if maxi < 1:
shapes[i] = [maxi, 1]
elif mini > 1:
shapes[i] = [1, 1 / mini]
self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride
# Cache images into memory for faster training (WARNING: large datasets may exceed system RAM)
self.imgs = [None] * n
if cache_images:
gb = 0 # Gigabytes of cached images
self.img_hw0, self.img_hw = [None] * n, [None] * n
results = ThreadPool(8).imap(lambda x: load_image(*x), zip(repeat(self), range(n))) # 8 threads
pbar = tqdm(enumerate(results), total=n)
for i, x in pbar:
self.imgs[i], self.img_hw0[i], self.img_hw[i] = x # img, hw_original, hw_resized = load_image(self, i)
gb += self.imgs[i].nbytes
pbar.desc = f'{prefix}Caching images ({gb / 1E9:.1f}GB)'
def cache_labels(self, path=Path('./labels.cache'), prefix=''):
# Cache dataset labels, check images and read shapes
x = {} # dict
nm, nf, ne, nc = 0, 0, 0, 0 # number missing, found, empty, duplicate
pbar = tqdm(zip(self.img_files, self.label_files), desc='Scanning images', total=len(self.img_files))
for i, (im_file, lb_file) in enumerate(pbar):
try:
# verify images
im = Image.open(im_file)
im.verify() # PIL verify
shape = exif_size(im) # image size
assert (shape[0] > 9) & (shape[1] > 9), 'image size <10 pixels'
# verify labels
if os.path.isfile(lb_file):
nf += 1 # label found
with open(lb_file, 'r') as f:
l = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32) # labels
if len(l):
assert l.shape[1] == 5, 'labels require 5 columns each'
assert (l >= 0).all(), 'negative labels'
assert (l[:, 1:] <= 1).all(), 'non-normalized or out of bounds coordinate labels'
assert np.unique(l, axis=0).shape[0] == l.shape[0], 'duplicate labels'
else:
ne += 1 # label empty
l = np.zeros((0, 5), dtype=np.float32)
else:
nm += 1 # label missing
l = np.zeros((0, 5), dtype=np.float32)
x[im_file] = [l, shape]
except Exception as e:
nc += 1
print(f'{prefix}WARNING: Ignoring corrupted image and/or label {im_file}: {e}')
pbar.desc = f"{prefix}Scanning '{path.parent / path.stem}' for images and labels... " \
f"{nf} found, {nm} missing, {ne} empty, {nc} corrupted"
if nf == 0:
print(f'{prefix}WARNING: No labels found in {path}. See {help_url}')
x['hash'] = get_hash(self.label_files + self.img_files)
x['results'] = [nf, nm, ne, nc, i + 1]
torch.save(x, path) # save for next time
logging.info(f'{prefix}New cache created: {path}')
return x
def __len__(self):
return len(self.img_files)
# def __iter__(self):
# self.count = -1
# print('ran dataset iter')
# #self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF)
# return self
def __getitem__(self, index):
index = self.indices[index] # linear, shuffled, or image_weights
hyp = self.hyp
mosaic = self.mosaic and random.random() < hyp['mosaic']
if mosaic:
# Load mosaic
img, labels = load_mosaic(self, index)
shapes = None
# MixUp https://arxiv.org/pdf/1710.09412.pdf
if random.random() < hyp['mixup']:
img2, labels2 = load_mosaic(self, random.randint(0, self.n - 1))
r = np.random.beta(8.0, 8.0) # mixup ratio, alpha=beta=8.0
img = (img * r + img2 * (1 - r)).astype(np.uint8)
labels = np.concatenate((labels, labels2), 0)
else:
# Load image
img, (h0, w0), (h, w) = load_image(self, index)
# Letterbox
shape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size # final letterboxed shape
img, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment)
shapes = (h0, w0), ((h / h0, w / w0), pad) # for COCO mAP rescaling
labels = self.labels[index].copy()
if labels.size: # normalized xywh to pixel xyxy format
labels[:, 1:] = xywhn2xyxy(labels[:, 1:], ratio[0] * w, ratio[1] * h, padw=pad[0], padh=pad[1])
if self.augment:
# Augment imagespace
if not mosaic:
img, labels = random_perspective(img, labels,
degrees=hyp['degrees'],
translate=hyp['translate'],
scale=hyp['scale'],
shear=hyp['shear'],
perspective=hyp['perspective'])
# Augment colorspace
augment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v'])
# Apply cutouts
# if random.random() < 0.9:
# labels = cutout(img, labels)
nL = len(labels) # number of labels
if nL:
labels[:, 1:5] = xyxy2xywh(labels[:, 1:5]) # convert xyxy to xywh
labels[:, [2, 4]] /= img.shape[0] # normalized height 0-1
labels[:, [1, 3]] /= img.shape[1] # normalized width 0-1
if self.augment:
# flip up-down
if random.random() < hyp['flipud']:
img = np.flipud(img)
if nL:
labels[:, 2] = 1 - labels[:, 2]
# flip left-right
if random.random() < hyp['fliplr']:
img = np.fliplr(img)
if nL:
labels[:, 1] = 1 - labels[:, 1]
labels_out = torch.zeros((nL, 6))
if nL:
labels_out[:, 1:] = torch.from_numpy(labels)
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
return torch.from_numpy(img), labels_out, self.img_files[index], shapes
@staticmethod
def collate_fn(batch):
img, label, path, shapes = zip(*batch) # transposed
for i, l in enumerate(label):
l[:, 0] = i # add target image index for build_targets()
return torch.stack(img, 0), torch.cat(label, 0), path, shapes
@staticmethod
def collate_fn4(batch):
img, label, path, shapes = zip(*batch) # transposed
n = len(shapes) // 4
img4, label4, path4, shapes4 = [], [], path[:n], shapes[:n]
ho = torch.tensor([[0., 0, 0, 1, 0, 0]])
wo = torch.tensor([[0., 0, 1, 0, 0, 0]])
s = torch.tensor([[1, 1, .5, .5, .5, .5]]) # scale
for i in range(n): # zidane torch.zeros(16,3,720,1280) # BCHW
i *= 4
if random.random() < 0.5:
im = F.interpolate(img[i].unsqueeze(0).float(), scale_factor=2., mode='bilinear', align_corners=False)[
0].type(img[i].type())
l = label[i]
else:
im = torch.cat((torch.cat((img[i], img[i + 1]), 1), torch.cat((img[i + 2], img[i + 3]), 1)), 2)
l = torch.cat((label[i], label[i + 1] + ho, label[i + 2] + wo, label[i + 3] + ho + wo), 0) * s
img4.append(im)
label4.append(l)
for i, l in enumerate(label4):
l[:, 0] = i # add target image index for build_targets()
return torch.stack(img4, 0), torch.cat(label4, 0), path4, shapes4
# Ancillary functions --------------------------------------------------------------------------------------------------
def load_image(self, index):
# loads 1 image from dataset, returns img, original hw, resized hw
img = self.imgs[index]
if img is None: # not cached
path = self.img_files[index]
img = cv2.imread(path) # BGR
assert img is not None, 'Image Not Found ' + path
h0, w0 = img.shape[:2] # orig hw
r = self.img_size / max(h0, w0) # resize image to img_size
if r != 1: # always resize down, only resize up if training with augmentation
interp = cv2.INTER_AREA if r < 1 and not self.augment else cv2.INTER_LINEAR
img = cv2.resize(img, (int(w0 * r), int(h0 * r)), interpolation=interp)
return img, (h0, w0), img.shape[:2] # img, hw_original, hw_resized
else:
return self.imgs[index], self.img_hw0[index], self.img_hw[index] # img, hw_original, hw_resized
def augment_hsv(img, hgain=0.5, sgain=0.5, vgain=0.5):
r = np.random.uniform(-1, 1, 3) * [hgain, sgain, vgain] + 1 # random gains
hue, sat, val = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))
dtype = img.dtype # uint8
x = np.arange(0, 256, dtype=np.int16)
lut_hue = ((x * r[0]) % 180).astype(dtype)
lut_sat = np.clip(x * r[1], 0, 255).astype(dtype)
lut_val = np.clip(x * r[2], 0, 255).astype(dtype)
img_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val))).astype(dtype)
cv2.cvtColor(img_hsv, cv2.COLOR_HSV2BGR, dst=img) # no return needed
# Histogram equalization
# if random.random() < 0.2:
# for i in range(3):
# img[:, :, i] = cv2.equalizeHist(img[:, :, i])
def load_mosaic(self, index):
# loads images in a 4-mosaic
labels4 = []
s = self.img_size
yc, xc = [int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border] # mosaic center x, y
indices = [index] + [self.indices[random.randint(0, self.n - 1)] for _ in range(3)] # 3 additional image indices
for i, index in enumerate(indices):
# Load image
img, _, (h, w) = load_image(self, index)
# place img in img4
if i == 0: # top left
img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles
x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc # xmin, ymin, xmax, ymax (large image)
x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h # xmin, ymin, xmax, ymax (small image)
elif i == 1: # top right
x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), yc
x1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), h
elif i == 2: # bottom left
x1a, y1a, x2a, y2a = max(xc - w, 0), yc, xc, min(s * 2, yc + h)
x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h)
elif i == 3: # bottom right
x1a, y1a, x2a, y2a = xc, yc, min(xc + w, s * 2), min(s * 2, yc + h)
x1b, y1b, x2b, y2b = 0, 0, min(w, x2a - x1a), min(y2a - y1a, h)
img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax]
padw = x1a - x1b
padh = y1a - y1b
# Labels
labels = self.labels[index].copy()
if labels.size:
labels[:, 1:] = xywhn2xyxy(labels[:, 1:], w, h, padw, padh) # normalized xywh to pixel xyxy format
labels4.append(labels)
# Concat/clip labels
if len(labels4):
labels4 = np.concatenate(labels4, 0)
np.clip(labels4[:, 1:], 0, 2 * s, out=labels4[:, 1:]) # use with random_perspective
# img4, labels4 = replicate(img4, labels4) # replicate
# Augment
img4, labels4 = random_perspective(img4, labels4,
degrees=self.hyp['degrees'],
translate=self.hyp['translate'],
scale=self.hyp['scale'],
shear=self.hyp['shear'],
perspective=self.hyp['perspective'],
border=self.mosaic_border) # border to remove
return img4, labels4
def load_mosaic9(self, index):
# loads images in a 9-mosaic
labels9 = []
s = self.img_size
indices = [index] + [self.indices[random.randint(0, self.n - 1)] for _ in range(8)] # 8 additional image indices
for i, index in enumerate(indices):
# Load image
img, _, (h, w) = load_image(self, index)
# place img in img9
if i == 0: # center
img9 = np.full((s * 3, s * 3, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles
h0, w0 = h, w
c = s, s, s + w, s + h # xmin, ymin, xmax, ymax (base) coordinates
elif i == 1: # top
c = s, s - h, s + w, s
elif i == 2: # top right
c = s + wp, s - h, s + wp + w, s
elif i == 3: # right
c = s + w0, s, s + w0 + w, s + h
elif i == 4: # bottom right
c = s + w0, s + hp, s + w0 + w, s + hp + h
elif i == 5: # bottom
c = s + w0 - w, s + h0, s + w0, s + h0 + h
elif i == 6: # bottom left
c = s + w0 - wp - w, s + h0, s + w0 - wp, s + h0 + h
elif i == 7: # left
c = s - w, s + h0 - h, s, s + h0
elif i == 8: # top left
c = s - w, s + h0 - hp - h, s, s + h0 - hp
padx, pady = c[:2]
x1, y1, x2, y2 = [max(x, 0) for x in c] # allocate coords
# Labels
labels = self.labels[index].copy()
if labels.size:
labels[:, 1:] = xywhn2xyxy(labels[:, 1:], w, h, padx, pady) # normalized xywh to pixel xyxy format
labels9.append(labels)
# Image
img9[y1:y2, x1:x2] = img[y1 - pady:, x1 - padx:] # img9[ymin:ymax, xmin:xmax]
hp, wp = h, w # height, width previous
# Offset
yc, xc = [int(random.uniform(0, s)) for x in self.mosaic_border] # mosaic center x, y
img9 = img9[yc:yc + 2 * s, xc:xc + 2 * s]
# Concat/clip labels
if len(labels9):
labels9 = np.concatenate(labels9, 0)
labels9[:, [1, 3]] -= xc
labels9[:, [2, 4]] -= yc
np.clip(labels9[:, 1:], 0, 2 * s, out=labels9[:, 1:]) # use with random_perspective
# img9, labels9 = replicate(img9, labels9) # replicate
# Augment
img9, labels9 = random_perspective(img9, labels9,
degrees=self.hyp['degrees'],
translate=self.hyp['translate'],
scale=self.hyp['scale'],
shear=self.hyp['shear'],
perspective=self.hyp['perspective'],
border=self.mosaic_border) # border to remove
return img9, labels9
def replicate(img, labels):
# Replicate labels
h, w = img.shape[:2]
boxes = labels[:, 1:].astype(int)
x1, y1, x2, y2 = boxes.T
s = ((x2 - x1) + (y2 - y1)) / 2 # side length (pixels)
for i in s.argsort()[:round(s.size * 0.5)]: # smallest indices
x1b, y1b, x2b, y2b = boxes[i]
bh, bw = y2b - y1b, x2b - x1b
yc, xc = int(random.uniform(0, h - bh)), int(random.uniform(0, w - bw)) # offset x, y
x1a, y1a, x2a, y2a = [xc, yc, xc + bw, yc + bh]
img[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax]
labels = np.append(labels, [[labels[i, 0], x1a, y1a, x2a, y2a]], axis=0)
return img, labels
def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True):
# Resize image to a 32-pixel-multiple rectangle https://github.com/ultralytics/yolov3/issues/232
shape = img.shape[:2] # current shape [height, width]
if isinstance(new_shape, int):
new_shape = (new_shape, new_shape)
# Scale ratio (new / old)
r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
if not scaleup: # only scale down, do not scale up (for better test mAP)
r = min(r, 1.0)
# Compute padding
ratio = r, r # width, height ratios
new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1] # wh padding
if auto: # minimum rectangle
dw, dh = np.mod(dw, 64), np.mod(dh, 64) # wh padding
elif scaleFill: # stretch
dw, dh = 0.0, 0.0
new_unpad = (new_shape[1], new_shape[0])
ratio = new_shape[1] / shape[1], new_shape[0] / shape[0] # width, height ratios
dw /= 2 # divide padding into 2 sides
dh /= 2
if shape[::-1] != new_unpad: # resize
img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)
top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color) # add border
return img, ratio, (dw, dh)
def random_perspective(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0, border=(0, 0)):
# torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))
# targets = [cls, xyxy]
height = img.shape[0] + border[0] * 2 # shape(h,w,c)
width = img.shape[1] + border[1] * 2
# Center
C = np.eye(3)
C[0, 2] = -img.shape[1] / 2 # x translation (pixels)
C[1, 2] = -img.shape[0] / 2 # y translation (pixels)
# Perspective
P = np.eye(3)
P[2, 0] = random.uniform(-perspective, perspective) # x perspective (about y)
P[2, 1] = random.uniform(-perspective, perspective) # y perspective (about x)
# Rotation and Scale
R = np.eye(3)
a = random.uniform(-degrees, degrees)
# a += random.choice([-180, -90, 0, 90]) # add 90deg rotations to small rotations
s = random.uniform(1 - scale, 1 + scale)
# s = 2 ** random.uniform(-scale, scale)
R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)
# Shear
S = np.eye(3)
S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # x shear (deg)
S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # y shear (deg)
# Translation
T = np.eye(3)
T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width # x translation (pixels)
T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height # y translation (pixels)
# Combined rotation matrix
M = T @ S @ R @ P @ C # order of operations (right to left) is IMPORTANT
if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any(): # image changed
if perspective:
img = cv2.warpPerspective(img, M, dsize=(width, height), borderValue=(114, 114, 114))
else: # affine
img = cv2.warpAffine(img, M[:2], dsize=(width, height), borderValue=(114, 114, 114))
# Visualize
# import matplotlib.pyplot as plt
# ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel()
# ax[0].imshow(img[:, :, ::-1]) # base
# ax[1].imshow(img2[:, :, ::-1]) # warped
# Transform label coordinates
n = len(targets)
if n:
# warp points
xy = np.ones((n * 4, 3))
xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2]].reshape(n * 4, 2) # x1y1, x2y2, x1y2, x2y1
xy = xy @ M.T # transform
if perspective:
xy = (xy[:, :2] / xy[:, 2:3]).reshape(n, 8) # rescale
else: # affine
xy = xy[:, :2].reshape(n, 8)
# create new boxes
x = xy[:, [0, 2, 4, 6]]
y = xy[:, [1, 3, 5, 7]]
xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T
# # apply angle-based reduction of bounding boxes
# radians = a * math.pi / 180
# reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5
# x = (xy[:, 2] + xy[:, 0]) / 2
# y = (xy[:, 3] + xy[:, 1]) / 2
# w = (xy[:, 2] - xy[:, 0]) * reduction
# h = (xy[:, 3] - xy[:, 1]) * reduction
# xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, n).T
# clip boxes
xy[:, [0, 2]] = xy[:, [0, 2]].clip(0, width)
xy[:, [1, 3]] = xy[:, [1, 3]].clip(0, height)
# filter candidates
i = box_candidates(box1=targets[:, 1:5].T * s, box2=xy.T)
targets = targets[i]
targets[:, 1:5] = xy[i]
return img, targets
def box_candidates(box1, box2, wh_thr=2, ar_thr=20, area_thr=0.1, eps=1e-16): # box1(4,n), box2(4,n)
# Compute candidate boxes: box1 before augment, box2 after augment, wh_thr (pixels), aspect_ratio_thr, area_ratio
w1, h1 = box1[2] - box1[0], box1[3] - box1[1]
w2, h2 = box2[2] - box2[0], box2[3] - box2[1]
ar = np.maximum(w2 / (h2 + eps), h2 / (w2 + eps)) # aspect ratio
return (w2 > wh_thr) & (h2 > wh_thr) & (w2 * h2 / (w1 * h1 + eps) > area_thr) & (ar < ar_thr) # candidates
def cutout(image, labels):
# Applies image cutout augmentation https://arxiv.org/abs/1708.04552
h, w = image.shape[:2]
def bbox_ioa(box1, box2):
# Returns the intersection over box2 area given box1, box2. box1 is 4, box2 is nx4. boxes are x1y1x2y2
box2 = box2.transpose()
# Get the coordinates of bounding boxes
b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3]
b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3]
# Intersection area
inter_area = (np.minimum(b1_x2, b2_x2) - np.maximum(b1_x1, b2_x1)).clip(0) * \
(np.minimum(b1_y2, b2_y2) - np.maximum(b1_y1, b2_y1)).clip(0)
# box2 area
box2_area = (b2_x2 - b2_x1) * (b2_y2 - b2_y1) + 1e-16
# Intersection over box2 area
return inter_area / box2_area
# create random masks
scales = [0.5] * 1 + [0.25] * 2 + [0.125] * 4 + [0.0625] * 8 + [0.03125] * 16 # image size fraction
for s in scales:
mask_h = random.randint(1, int(h * s))
mask_w = random.randint(1, int(w * s))
# box
xmin = max(0, random.randint(0, w) - mask_w // 2)
ymin = max(0, random.randint(0, h) - mask_h // 2)
xmax = min(w, xmin + mask_w)
ymax = min(h, ymin + mask_h)
# apply random color mask
image[ymin:ymax, xmin:xmax] = [random.randint(64, 191) for _ in range(3)]
# return unobscured labels
if len(labels) and s > 0.03:
box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32)
ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area
labels = labels[ioa < 0.60] # remove >60% obscured labels
return labels
def create_folder(path='./new'):
# Create folder
if os.path.exists(path):
shutil.rmtree(path) # delete output folder
os.makedirs(path) # make new output folder
def flatten_recursive(path='../coco128'):
# Flatten a recursive directory by bringing all files to top level
new_path = Path(path + '_flat')
create_folder(new_path)
for file in tqdm(glob.glob(str(Path(path)) + '/**/*.*', recursive=True)):
shutil.copyfile(file, new_path / Path(file).name)
def extract_boxes(path='../coco128/'): # from utils.datasets import *; extract_boxes('../coco128')
# Convert detection dataset into classification dataset, with one directory per class
path = Path(path) # images dir
shutil.rmtree(path / 'classifier') if (path / 'classifier').is_dir() else None # remove existing
files = list(path.rglob('*.*'))
n = len(files) # number of files
for im_file in tqdm(files, total=n):
if im_file.suffix[1:] in img_formats:
# image
im = cv2.imread(str(im_file))[..., ::-1] # BGR to RGB
h, w = im.shape[:2]
# labels
lb_file = Path(img2label_paths([str(im_file)])[0])
if Path(lb_file).exists():
with open(lb_file, 'r') as f:
lb = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32) # labels
for j, x in enumerate(lb):
c = int(x[0]) # class
f = (path / 'classifier') / f'{c}' / f'{path.stem}_{im_file.stem}_{j}.jpg' # new filename
if not f.parent.is_dir():
f.parent.mkdir(parents=True)
b = x[1:] * [w, h, w, h] # box
# b[2:] = b[2:].max() # rectangle to square
b[2:] = b[2:] * 1.2 + 3 # pad
b = xywh2xyxy(b.reshape(-1, 4)).ravel().astype(np.int)
b[[0, 2]] = np.clip(b[[0, 2]], 0, w) # clip boxes outside of image
b[[1, 3]] = np.clip(b[[1, 3]], 0, h)
assert cv2.imwrite(str(f), im[b[1]:b[3], b[0]:b[2]]), f'box failure in {f}'
def autosplit(path='../coco128', weights=(0.9, 0.1, 0.0)): # from utils.datasets import *; autosplit('../coco128')
""" Autosplit a dataset into train/val/test splits and save path/autosplit_*.txt files
# Arguments
path: Path to images directory
weights: Train, val, test weights (list)
"""
path = Path(path) # images dir
files = list(path.rglob('*.*'))
n = len(files) # number of files
indices = random.choices([0, 1, 2], weights=weights, k=n) # assign each image to a split
txt = ['autosplit_train.txt', 'autosplit_val.txt', 'autosplit_test.txt'] # 3 txt files
[(path / x).unlink() for x in txt if (path / x).exists()] # remove existing
for i, img in tqdm(zip(indices, files), total=n):
if img.suffix[1:] in img_formats:
with open(path / txt[i], 'a') as f:
f.write(str(img) + '\n') # add image to txt file
================================================
FILE: utils/face_datasets.py
================================================
import glob
import logging
import math
import os
import random
import shutil
import time
from itertools import repeat
from multiprocessing.pool import ThreadPool
from pathlib import Path
from threading import Thread
import cv2
import numpy as np
import torch
from PIL import Image, ExifTags
from torch.utils.data import Dataset
from tqdm import tqdm
from utils.general import xyxy2xywh, xywh2xyxy, clean_str
from utils.torch_utils import torch_distributed_zero_first
# Parameters
help_url = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
img_formats = ['bmp', 'jpg', 'jpeg', 'png', 'tif', 'tiff', 'dng'] # acceptable image suffixes
vid_formats = ['mov', 'avi', 'mp4', 'mpg', 'mpeg', 'm4v', 'wmv', 'mkv'] # acceptable video suffixes
logger = logging.getLogger(__name__)
# Get orientation exif tag
for orientation in ExifTags.TAGS.keys():
if ExifTags.TAGS[orientation] == 'Orientation':
break
def get_hash(files):
# Returns a single hash value of a list of files
return sum(os.path.getsize(f) for f in files if os.path.isfile(f))
def img2label_paths(img_paths):
# Define label paths as a function of image paths
sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep # /images/, /labels/ substrings
return [x.replace(sa, sb, 1).replace('.' + x.split('.')[-1], '.txt') for x in img_paths]
def exif_size(img):
# Returns exif-corrected PIL size
s = img.size # (width, height)
try:
rotation = dict(img._getexif().items())[orientation]
if rotation == 6: # rotation 270
s = (s[1], s[0])
elif rotation == 8: # rotation 90
s = (s[1], s[0])
except:
pass
return s
def create_dataloader(path, imgsz, batch_size, stride, opt, hyp=None, augment=False, cache=False, pad=0.0, rect=False,
rank=-1, world_size=1, workers=8, image_weights=False, quad=False, prefix=''):
# Make sure only the first process in DDP process the dataset first, and the following others can use the cache
with torch_distributed_zero_first(rank):
dataset = LoadFaceImagesAndLabels(path, imgsz, batch_size,
augment=augment, # augment images
hyp=hyp, # augmentation hyperparameters
rect=rect, # rectangular training
cache_images=cache,
single_cls=opt.single_cls,
stride=int(stride),
pad=pad,
image_weights=image_weights,
)
batch_size = min(batch_size, len(dataset))
nw = min([os.cpu_count() // world_size, batch_size if batch_size > 1 else 0, workers]) # number of workers
sampler = torch.utils.data.distributed.DistributedSampler(dataset) if rank != -1 else None
loader = torch.utils.data.DataLoader if image_weights else InfiniteDataLoader
# Use torch.utils.data.DataLoader() if dataset.properties will update during training else InfiniteDataLoader()
dataloader = loader(dataset,
batch_size=batch_size,
num_workers=nw,
sampler=sampler,
pin_memory=True,
collate_fn=LoadFaceImagesAndLabels.collate_fn4 if quad else LoadFaceImagesAndLabels.collate_fn)
return dataloader, dataset
class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):
""" Dataloader that reuses workers
Uses same syntax as vanilla DataLoader
"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
object.__setattr__(self, 'batch_sampler', _RepeatSampler(self.batch_sampler))
self.iterator = super().__iter__()
def __len__(self):
return len(self.batch_sampler.sampler)
def __iter__(self):
for i in range(len(self)):
yield next(self.iterator)
class _RepeatSampler(object):
""" Sampler that repeats forever
Args:
sampler (Sampler)
"""
def __init__(self, sampler):
self.sampler = sampler
def __iter__(self):
while True:
yield from iter(self.sampler)
class LoadFaceImagesAndLabels(Dataset): # for training/testing
def __init__(self, path, img_size=640, batch_size=16, augment=False, hyp=None, rect=False, image_weights=False,
cache_images=False, single_cls=False, stride=32, pad=0.0, rank=-1):
self.img_size = img_size
self.augment = augment
self.hyp = hyp
self.image_weights = image_weights
self.rect = False if image_weights else rect
self.mosaic = self.augment and not self.rect # load 4 images at a time into a mosaic (only during training)
self.mosaic_border = [-img_size // 2, -img_size // 2]
self.stride = stride
try:
f = [] # image files
for p in path if isinstance(path, list) else [path]:
p = Path(p) # os-agnostic
if p.is_dir(): # dir
f += glob.glob(str(p / '**' / '*.*'), recursive=True)
elif p.is_file(): # file
with open(p, 'r') as t:
t = t.read().strip().splitlines()
parent = str(p.parent) + os.sep
f += [x.replace('./', parent) if x.startswith('./') else x for x in t] # local to global path
else:
raise Exception('%s does not exist' % p)
self.img_files = sorted([x.replace('/', os.sep) for x in f if x.split('.')[-1].lower() in img_formats])
assert self.img_files, 'No images found'
except Exception as e:
raise Exception('Error loading data from %s: %s\nSee %s' % (path, e, help_url))
# Check cache
self.label_files = img2label_paths(self.img_files) # labels
cache_path = Path(self.label_files[0]).parent.with_suffix('.cache') # cached labels
if cache_path.is_file():
cache = torch.load(cache_path) # load
if cache['hash'] != get_hash(self.label_files + self.img_files) or 'results' not in cache: # changed
cache = self.cache_labels(cache_path) # re-cache
else:
cache = self.cache_labels(cache_path) # cache
# Display cache
[nf, nm, ne, nc, n] = cache.pop('results') # found, missing, empty, corrupted, total
desc = f"Scanning '{cache_path}' for images and labels... {nf} found, {nm} missing, {ne} empty, {nc} corrupted"
tqdm(None, desc=desc, total=n, initial=n)
assert nf > 0 or not augment, f'No labels found in {cache_path}. Can not train without labels. See {help_url}'
# Read cache
cache.pop('hash') # remove hash
labels, shapes = zip(*cache.values())
self.labels = list(labels)
self.shapes = np.array(shapes, dtype=np.float64)
self.img_files = list(cache.keys()) # update
self.label_files = img2label_paths(cache.keys()) # update
if single_cls:
for x in self.labels:
x[:, 0] = 0
n = len(shapes) # number of images
bi = np.floor(np.arange(n) / batch_size).astype(np.int) # batch index
nb = bi[-1] + 1 # number of batches
self.batch = bi # batch index of image
self.n = n
self.indices = range(n)
# Rectangular Training
if self.rect:
# Sort by aspect ratio
s = self.shapes # wh
ar = s[:, 1] / s[:, 0] # aspect ratio
irect = ar.argsort()
self.img_files = [self.img_files[i] for i in irect]
self.label_files = [self.label_files[i] for i in irect]
self.labels = [self.labels[i] for i in irect]
self.shapes = s[irect] # wh
ar = ar[irect]
# Set training image shapes
shapes = [[1, 1]] * nb
for i in range(nb):
ari = ar[bi == i]
mini, maxi = ari.min(), ari.max()
if maxi < 1:
shapes[i] = [maxi, 1]
elif mini > 1:
shapes[i] = [1, 1 / mini]
self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride
# Cache images into memory for faster training (WARNING: large datasets may exceed system RAM)
self.imgs = [None] * n
if cache_images:
gb = 0 # Gigabytes of cached images
self.img_hw0, self.img_hw = [None] * n, [None] * n
results = ThreadPool(8).imap(lambda x: load_image(*x), zip(repeat(self), range(n))) # 8 threads
pbar = tqdm(enumerate(results), total=n)
for i, x in pbar:
self.imgs[i], self.img_hw0[i], self.img_hw[i] = x # img, hw_original, hw_resized = load_image(self, i)
gb += self.imgs[i].nbytes
pbar.desc = 'Caching images (%.1fGB)' % (gb / 1E9)
def cache_labels(self, path=Path('./labels.cache')):
# Cache dataset labels, check images and read shapes
x = {} # dict
nm, nf, ne, nc = 0, 0, 0, 0 # number missing, found, empty, duplicate
pbar = tqdm(zip(self.img_files, self.label_files), desc='Scanning images', total=len(self.img_files))
for i, (im_file, lb_file) in enumerate(pbar):
try:
# verify images
im = Image.open(im_file)
im.verify() # PIL verify
shape = exif_size(im) # image size
assert (shape[0] > 9) & (shape[1] > 9), 'image size <10 pixels'
# verify labels
if os.path.isfile(lb_file):
nf += 1 # label found
with open(lb_file, 'r') as f:
l = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32) # labels
if len(l):
assert l.shape[1] == 15, 'labels require 15 columns each'
assert (l >= -1).all(), 'negative labels'
assert (l[:, 1:] <= 1).all(), 'non-normalized or out of bounds coordinate labels'
assert np.unique(l, axis=0).shape[0] == l.shape[0], 'duplicate labels'
else:
ne += 1 # label empty
l = np.zeros((0, 15), dtype=np.float32)
else:
nm += 1 # label missing
l = np.zeros((0, 15), dtype=np.float32)
x[im_file] = [l, shape]
except Exception as e:
nc += 1
print('WARNING: Ignoring corrupted image and/or label %s: %s' % (im_file, e))
pbar.desc = f"Scanning '{path.parent / path.stem}' for images and labels... " \
f"{nf} found, {nm} missing, {ne} empty, {nc} corrupted"
if nf == 0:
print(f'WARNING: No labels found in {path}. See {help_url}')
x['hash'] = get_hash(self.label_files + self.img_files)
x['results'] = [nf, nm, ne, nc, i + 1]
torch.save(x, path) # save for next time
logging.info(f"New cache created: {path}")
return x
def __len__(self):
return len(self.img_files)
# def __iter__(self):
# self.count = -1
# print('ran dataset iter')
# #self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF)
# return self
def __getitem__(self, index):
index = self.indices[index] # linear, shuffled, or image_weights
hyp = self.hyp
mosaic = self.mosaic and random.random() < hyp['mosaic']
if mosaic:
# Load mosaic
img, labels = load_mosaic_face(self, index)
shapes = None
# MixUp https://arxiv.org/pdf/1710.09412.pdf
if random.random() < hyp['mixup']:
img2, labels2 = load_mosaic_face(self, random.randint(0, self.n - 1))
r = np.random.beta(8.0, 8.0) # mixup ratio, alpha=beta=8.0
img = (img * r + img2 * (1 - r)).astype(np.uint8)
labels = np.concatenate((labels, labels2), 0)
else:
# Load image
img, (h0, w0), (h, w) = load_image(self, index)
# Letterbox
shape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size # final letterboxed shape
img, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment)
shapes = (h0, w0), ((h / h0, w / w0), pad) # for COCO mAP rescaling
# Load labels
labels = []
x = self.labels[index]
if x.size > 0:
# Normalized xywh to pixel xyxy format
labels = x.copy()
labels[:, 1] = ratio[0] * w * (x[:, 1] - x[:, 3] / 2) + pad[0] # pad width
labels[:, 2] = ratio[1] * h * (x[:, 2] - x[:, 4] / 2) + pad[1] # pad height
labels[:, 3] = ratio[0] * w * (x[:, 1] + x[:, 3] / 2) + pad[0]
labels[:, 4] = ratio[1] * h * (x[:, 2] + x[:, 4] / 2) + pad[1]
#labels[:, 5] = ratio[0] * w * x[:, 5] + pad[0] # pad width
labels[:, 5] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 5] + pad[0]) + (
np.array(x[:, 5] > 0, dtype=np.int32) - 1)
labels[:, 6] = np.array(x[:, 6] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 6] + pad[1]) + (
np.array(x[:, 6] > 0, dtype=np.int32) - 1)
labels[:, 7] = np.array(x[:, 7] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 7] + pad[0]) + (
np.array(x[:, 7] > 0, dtype=np.int32) - 1)
labels[:, 8] = np.array(x[:, 8] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 8] + pad[1]) + (
np.array(x[:, 8] > 0, dtype=np.int32) - 1)
labels[:, 9] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 9] + pad[0]) + (
np.array(x[:, 9] > 0, dtype=np.int32) - 1)
labels[:, 10] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 10] + pad[1]) + (
np.array(x[:, 10] > 0, dtype=np.int32) - 1)
labels[:, 11] = np.array(x[:, 11] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 11] + pad[0]) + (
np.array(x[:, 11] > 0, dtype=np.int32) - 1)
labels[:, 12] = np.array(x[:, 12] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 12] + pad[1]) + (
np.array(x[:, 12] > 0, dtype=np.int32) - 1)
labels[:, 13] = np.array(x[:, 13] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 13] + pad[0]) + (
np.array(x[:, 13] > 0, dtype=np.int32) - 1)
labels[:, 14] = np.array(x[:, 14] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 14] + pad[1]) + (
np.array(x[:, 14] > 0, dtype=np.int32) - 1)
if self.augment:
# Augment imagespace
if not mosaic:
img, labels = random_perspective(img, labels,
degrees=hyp['degrees'],
translate=hyp['translate'],
scale=hyp['scale'],
shear=hyp['shear'],
perspective=hyp['perspective'])
# Augment colorspace
augment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v'])
# Apply cutouts
# if random.random() < 0.9:
# labels = cutout(img, labels)
nL = len(labels) # number of labels
if nL:
labels[:, 1:5] = xyxy2xywh(labels[:, 1:5]) # convert xyxy to xywh
labels[:, [2, 4]] /= img.shape[0] # normalized height 0-1
labels[:, [1, 3]] /= img.shape[1] # normalized width 0-1
labels[:, [5, 7, 9, 11, 13]] /= img.shape[1] # normalized landmark x 0-1
labels[:, [5, 7, 9, 11, 13]] = np.where(labels[:, [5, 7, 9, 11, 13]] < 0, -1, labels[:, [5, 7, 9, 11, 13]])
labels[:, [6, 8, 10, 12, 14]] /= img.shape[0] # normalized landmark y 0-1
labels[:, [6, 8, 10, 12, 14]] = np.where(labels[:, [6, 8, 10, 12, 14]] < 0, -1, labels[:, [6, 8, 10, 12, 14]])
if self.augment:
# flip up-down
if random.random() < hyp['flipud']:
img = np.flipud(img)
if nL:
labels[:, 2] = 1 - labels[:, 2]
labels[:, 6] = np.where(labels[:,6] < 0, -1, 1 - labels[:, 6])
labels[:, 8] = np.where(labels[:, 8] < 0, -1, 1 - labels[:, 8])
labels[:, 10] = np.where(labels[:, 10] < 0, -1, 1 - labels[:, 10])
labels[:, 12] = np.where(labels[:, 12] < 0, -1, 1 - labels[:, 12])
labels[:, 14] = np.where(labels[:, 14] < 0, -1, 1 - labels[:, 14])
# flip left-right
if random.random() < hyp['fliplr']:
img = np.fliplr(img)
if nL:
labels[:, 1] = 1 - labels[:, 1]
labels[:, 5] = np.where(labels[:, 5] < 0, -1, 1 - labels[:, 5])
labels[:, 7] = np.where(labels[:, 7] < 0, -1, 1 - labels[:, 7])
labels[:, 9] = np.where(labels[:, 9] < 0, -1, 1 - labels[:, 9])
labels[:, 11] = np.where(labels[:, 11] < 0, -1, 1 - labels[:, 11])
labels[:, 13] = np.where(labels[:, 13] < 0, -1, 1 - labels[:, 13])
#左右镜像的时候，左眼、右眼， 左嘴角、右嘴角无法区分, 应该交换位置，便于网络学习
eye_left = np.copy(labels[:, [5, 6]])
mouth_left = np.copy(labels[:, [11, 12]])
labels[:, [5, 6]] = labels[:, [7, 8]]
labels[:, [7, 8]] = eye_left
labels[:, [11, 12]] = labels[:, [13, 14]]
labels[:, [13, 14]] = mouth_left
labels_out = torch.zeros((nL, 16))
if nL:
labels_out[:, 1:] = torch.from_numpy(labels)
#showlabels(img, labels[:, 1:5], labels[:, 5:15])
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
#print(index, ' --- labels_out: ', labels_out)
#if nL:
#print( ' : landmarks : ', torch.max(labels_out[:, 5:15]), ' --- ', torch.min(labels_out[:, 5:15]))
return torch.from_numpy(img), labels_out, self.img_files[index], shapes
@staticmethod
def collate_fn(batch):
img, label, path, shapes = zip(*batch) # transposed
for i, l in enumerate(label):
l[:, 0] = i # add target image index for build_targets()
return torch.stack(img, 0), torch.cat(label, 0), path, shapes
def showlabels(img, boxs, landmarks):
for box in boxs:
x,y,w,h = box[0] * img.shape[1], box[1] * img.shape[0], box[2] * img.shape[1], box[3] * img.shape[0]
#cv2.rectangle(image, (x,y), (x+w,y+h), (0,255,0), 2)
cv2.rectangle(img, (int(x - w/2), int(y - h/2)), (int(x + w/2), int(y + h/2)), (0, 255, 0), 2)
for landmark in landmarks:
#cv2.circle(img,(60,60),30,(0,0,255))
for i in range(5):
cv2.circle(img, (int(landmark[2*i] * img.shape[1]), int(landmark[2*i+1]*img.shape[0])), 3 ,(0,0,255), -1)
cv2.imshow('test', img)
cv2.waitKey(0)
def load_mosaic_face(self, index):
# loads images in a mosaic
labels4 = []
s = self.img_size
yc, xc = [int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border] # mosaic center x, y
indices = [index] + [self.indices[random.randint(0, self.n - 1)] for _ in range(3)] # 3 additional image indices
for i, index in enumerate(indices):
# Load image
img, _, (h, w) = load_image(self, index)
# place img in img4
if i == 0: # top left
img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles
x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc # xmin, ymin, xmax, ymax (large image)
x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h # xmin, ymin, xmax, ymax (small image)
elif i == 1: # top right
x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), yc
x1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), h
elif i == 2: # bottom left
x1a, y1a, x2a, y2a = max(xc - w, 0), yc, xc, min(s * 2, yc + h)
x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h)
elif i == 3: # bottom right
x1a, y1a, x2a, y2a = xc, yc, min(xc + w, s * 2), min(s * 2, yc + h)
x1b, y1b, x2b, y2b = 0, 0, min(w, x2a - x1a), min(y2a - y1a, h)
img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax]
padw = x1a - x1b
padh = y1a - y1b
# Labels
x = self.labels[index]
labels = x.copy()
if x.size > 0: # Normalized xywh to pixel xyxy format
#box, x1,y1,x2,y2
labels[:, 1] = w * (x[:, 1] - x[:, 3] / 2) + padw
labels[:, 2] = h * (x[:, 2] - x[:, 4] / 2) + padh
labels[:, 3] = w * (x[:, 1] + x[:, 3] / 2) + padw
labels[:, 4] = h * (x[:, 2] + x[:, 4] / 2) + padh
#10 landmarks
labels[:, 5] = np.array(x[:, 5] > 0, dtype=np.int32) * (w * x[:, 5] + padw) + (np.array(x[:, 5] > 0, dtype=np.int32) - 1)
labels[:, 6] = np.array(x[:, 6] > 0, dtype=np.int32) * (h * x[:, 6] + padh) + (np.array(x[:, 6] > 0, dtype=np.int32) - 1)
labels[:, 7] = np.array(x[:, 7] > 0, dtype=np.int32) * (w * x[:, 7] + padw) + (np.array(x[:, 7] > 0, dtype=np.int32) - 1)
labels[:, 8] = np.array(x[:, 8] > 0, dtype=np.int32) * (h * x[:, 8] + padh) + (np.array(x[:, 8] > 0, dtype=np.int32) - 1)
labels[:, 9] = np.array(x[:, 9] > 0, dtype=np.int32) * (w * x[:, 9] + padw) + (np.array(x[:, 9] > 0, dtype=np.int32) - 1)
labels[:, 10] = np.array(x[:, 10] > 0, dtype=np.int32) * (h * x[:, 10] + padh) + (np.array(x[:, 10] > 0, dtype=np.int32) - 1)
labels[:, 11] = np.array(x[:, 11] > 0, dtype=np.int32) * (w * x[:, 11] + padw) + (np.array(x[:, 11] > 0, dtype=np.int32) - 1)
labels[:, 12] = np.array(x[:, 12] > 0, dtype=np.int32) * (h * x[:, 12] + padh) + (np.array(x[:, 12] > 0, dtype=np.int32) - 1)
labels[:, 13] = np.array(x[:, 13] > 0, dtype=np.int32) * (w * x[:, 13] + padw) + (np.array(x[:, 13] > 0, dtype=np.int32) - 1)
labels[:, 14] = np.array(x[:, 14] > 0, dtype=np.int32) * (h * x[:, 14] + padh) + (np.array(x[:, 14] > 0, dtype=np.int32) - 1)
labels4.append(labels)
# Concat/clip labels
if len(labels4):
labels4 = np.concatenate(labels4, 0)
np.clip(labels4[:, 1:5], 0, 2 * s, out=labels4[:, 1:5]) # use with random_perspective
# img4, labels4 = replicate(img4, labels4) # replicate
#landmarks
labels4[:, 5:] = np.where(labels4[:, 5:] < 0, -1, labels4[:, 5:])
labels4[:, 5:] = np.where(labels4[:, 5:] > 2 * s, -1, labels4[:, 5:])
labels4[:, 5] = np.where(labels4[:, 6] == -1, -1, labels4[:, 5])
labels4[:, 6] = np.where(labels4[:, 5] == -1, -1, labels4[:, 6])
labels4[:, 7] = np.where(labels4[:, 8] == -1, -1, labels4[:, 7])
labels4[:, 8] = np.where(labels4[:, 7] == -1, -1, labels4[:, 8])
labels4[:, 9] = np.where(labels4[:, 10] == -1, -1, labels4[:, 9])
labels4[:, 10] = np.where(labels4[:, 9] == -1, -1, labels4[:, 10])
labels4[:, 11] = np.where(labels4[:, 12] == -1, -1, labels4[:, 11])
labels4[:, 12] = np.where(labels4[:, 11] == -1, -1, labels4[:, 12])
labels4[:, 13] = np.where(labels4[:, 14] == -1, -1, labels4[:, 13])
labels4[:, 14] = np.where(labels4[:, 13] == -1, -1, labels4[:, 14])
# Augment
img4, labels4 = random_perspective(img4, labels4,
degrees=self.hyp['degrees'],
translate=self.hyp['translate'],
scale=self.hyp['scale'],
shear=self.hyp['shear'],
perspective=self.hyp['perspective'],
border=self.mosaic_border) # border to remove
return img4, labels4
# Ancillary functions --------------------------------------------------------------------------------------------------
def load_image(self, index):
# loads 1 image from dataset, returns img, original hw, resized hw
img = self.imgs[index]
if img is None: # not cached
path = self.img_files[index]
img = cv2.imread(path) # BGR
assert img is not None, 'Image Not Found ' + path
h0, w0 = img.shape[:2] # orig hw
r = self.img_size / max(h0, w0) # resize image to img_size
if r != 1: # always resize down, only resize up if training with augmentation
interp = cv2.INTER_AREA if r < 1 and not self.augment else cv2.INTER_LINEAR
img = cv2.resize(img, (int(w0 * r), int(h0 * r)), interpolation=interp)
return img, (h0, w0), img.shape[:2] # img, hw_original, hw_resized
else:
return self.imgs[index], self.img_hw0[index], self.img_hw[index] # img, hw_original, hw_resized
def augment_hsv(img, hgain=0.5, sgain=0.5, vgain=0.5):
r = np.random.uniform(-1, 1, 3) * [hgain, sgain, vgain] + 1 # random gains
hue, sat, val = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))
dtype = img.dtype # uint8
x = np.arange(0, 256, dtype=np.int16)
lut_hue = ((x * r[0]) % 180).astype(dtype)
lut_sat = np.clip(x * r[1], 0, 255).astype(dtype)
lut_val = np.clip(x * r[2], 0, 255).astype(dtype)
img_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val))).astype(dtype)
cv2.cvtColor(img_hsv, cv2.COLOR_HSV2BGR, dst=img) # no return needed
# Histogram equalization
# if random.random() < 0.2:
# for i in range(3):
# img[:, :, i] = cv2.equalizeHist(img[:, :, i])
def replicate(img, labels):
# Replicate labels
h, w = img.shape[:2]
boxes = labels[:, 1:].astype(int)
x1, y1, x2, y2 = boxes.T
s = ((x2 - x1) + (y2 - y1)) / 2 # side length (pixels)
for i in s.argsort()[:round(s.size * 0.5)]: # smallest indices
x1b, y1b, x2b, y2b = boxes[i]
bh, bw = y2b - y1b, x2b - x1b
yc, xc = int(random.uniform(0, h - bh)), int(random.uniform(0, w - bw)) # offset x, y
x1a, y1a, x2a, y2a = [xc, yc, xc + bw, yc + bh]
img[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax]
labels = np.append(labels, [[labels[i, 0], x1a, y1a, x2a, y2a]], axis=0)
return img, labels
def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True):
# Resize image to a 32-pixel-multiple rectangle https://github.com/ultralytics/yolov3/issues/232
shape = img.shape[:2] # current shape [height, width]
if isinstance(new_shape, int):
new_shape = (new_shape, new_shape)
# Scale ratio (new / old)
r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
if not scaleup: # only scale down, do not scale up (for better test mAP)
r = min(r, 1.0)
# Compute padding
ratio = r, r # width, height ratios
new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1] # wh padding
if auto: # minimum rectangle
dw, dh = np.mod(dw, 64), np.mod(dh, 64) # wh padding
elif scaleFill: # stretch
dw, dh = 0.0, 0.0
new_unpad = (new_shape[1], new_shape[0])
ratio = new_shape[1] / shape[1], new_shape[0] / shape[0] # width, height ratios
dw /= 2 # divide padding into 2 sides
dh /= 2
if shape[::-1] != new_unpad: # resize
img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)
top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color) # add border
return img, ratio, (dw, dh)
def random_perspective(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0, border=(0, 0)):
# torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))
# targets = [cls, xyxy]
height = img.shape[0] + border[0] * 2 # shape(h,w,c)
width = img.shape[1] + border[1] * 2
# Center
C = np.eye(3)
C[0, 2] = -img.shape[1] / 2 # x translation (pixels)
C[1, 2] = -img.shape[0] / 2 # y translation (pixels)
# Perspective
P = np.eye(3)
P[2, 0] = random.uniform(-perspective, perspective) # x perspective (about y)
P[2, 1] = random.uniform(-perspective, perspective) # y perspective (about x)
# Rotation and Scale
R = np.eye(3)
a = random.uniform(-degrees, degrees)
# a += random.choice([-180, -90, 0, 90]) # add 90deg rotations to small rotations
s = random.uniform(1 - scale, 1 + scale)
# s = 2 ** random.uniform(-scale, scale)
R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)
# Shear
S = np.eye(3)
S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # x shear (deg)
S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # y shear (deg)
# Translation
T = np.eye(3)
T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width # x translation (pixels)
T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height # y translation (pixels)
# Combined rotation matrix
M = T @ S @ R @ P @ C # order of operations (right to left) is IMPORTANT
if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any(): # image changed
if perspective:
img = cv2.warpPerspective(img, M, dsize=(width, height), borderValue=(114, 114, 114))
else: # affine
img = cv2.warpAffine(img, M[:2], dsize=(width, height), borderValue=(114, 114, 114))
# Visualize
# import matplotlib.pyplot as plt
# ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel()
# ax[0].imshow(img[:, :, ::-1]) # base
# ax[1].imshow(img2[:, :, ::-1]) # warped
# Transform label coordinates
n = len(targets)
if n:
# warp points
#xy = np.ones((n * 4, 3))
xy = np.ones((n * 9, 3))
xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]].reshape(n * 9, 2) # x1y1, x2y2, x1y2, x2y1
xy = xy @ M.T # transform
if perspective:
xy = (xy[:, :2] / xy[:, 2:3]).reshape(n, 18) # rescale
else: # affine
xy = xy[:, :2].reshape(n, 18)
# create new boxes
x = xy[:, [0, 2, 4, 6]]
y = xy[:, [1, 3, 5, 7]]
landmarks = xy[:, [8, 9, 10, 11, 12, 13, 14, 15, 16, 17]]
mask = np.array(targets[:, 5:] > 0, dtype=np.int32)
landmarks = landmarks * mask
landmarks = landmarks + mask - 1
landmarks = np.where(landmarks < 0, -1, landmarks)
landmarks[:, [0, 2, 4, 6, 8]] = np.where(landmarks[:, [0, 2, 4, 6, 8]] > width, -1, landmarks[:, [0, 2, 4, 6, 8]])
landmarks[:, [1, 3, 5, 7, 9]] = np.where(landmarks[:, [1, 3, 5, 7, 9]] > height, -1,landmarks[:, [1, 3, 5, 7, 9]])
landmarks[:, 0] = np.where(landmarks[:, 1] == -1, -1, landmarks[:, 0])
landmarks[:, 1] = np.where(landmarks[:, 0] == -1, -1, landmarks[:, 1])
landmarks[:, 2] = np.where(landmarks[:, 3] == -1, -1, landmarks[:, 2])
landmarks[:, 3] = np.where(landmarks[:, 2] == -1, -1, landmarks[:, 3])
landmarks[:, 4] = np.where(landmarks[:, 5] == -1, -1, landmarks[:, 4])
landmarks[:, 5] = np.where(landmarks[:, 4] == -1, -1, landmarks[:, 5])
landmarks[:, 6] = np.where(landmarks[:, 7] == -1, -1, landmarks[:, 6])
landmarks[:, 7] = np.where(landmarks[:, 6] == -1, -1, landmarks[:, 7])
landmarks[:, 8] = np.where(landmarks[:, 9] == -1, -1, landmarks[:, 8])
landmarks[:, 9] = np.where(landmarks[:, 8] == -1, -1, landmarks[:, 9])
targets[:,5:] = landmarks
xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T
# # apply angle-based reduction of bounding boxes
# radians = a * math.pi / 180
# reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5
# x = (xy[:, 2] + xy[:, 0]) / 2
# y = (xy[:, 3] + xy[:, 1]) / 2
# w = (xy[:, 2] - xy[:, 0]) * reduction
# h = (xy[:, 3] - xy[:, 1]) * reduction
# xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, n).T
# clip boxes
xy[:, [0, 2]] = xy[:, [0, 2]].clip(0, width)
xy[:, [1, 3]] = xy[:, [1, 3]].clip(0, height)
# filter candidates
i = box_candidates(box1=targets[:, 1:5].T * s, box2=xy.T)
targets = targets[i]
targets[:, 1:5] = xy[i]
return img, targets
def box_candidates(box1, box2, wh_thr=2, ar_thr=20, area_thr=0.1): # box1(4,n), box2(4,n)
# Compute candidate boxes: box1 before augment, box2 after augment, wh_thr (pixels), aspect_ratio_thr, area_ratio
w1, h1 = box1[2] - box1[0], box1[3] - box1[1]
w2, h2 = box2[2] - box2[0], box2[3] - box2[1]
ar = np.maximum(w2 / (h2 + 1e-16), h2 / (w2 + 1e-16)) # aspect ratio
return (w2 > wh_thr) & (h2 > wh_thr) & (w2 * h2 / (w1 * h1 + 1e-16) > area_thr) & (ar < ar_thr) # candidates
def cutout(image, labels):
# Applies image cutout augmentation https://arxiv.org/abs/1708.04552
h, w = image.shape[:2]
def bbox_ioa(box1, box2):
# Returns the intersection over box2 area given box1, box2. box1 is 4, box2 is nx4. boxes are x1y1x2y2
box2 = box2.transpose()
# Get the coordinates of bounding boxes
b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3]
b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3]
# Intersection area
inter_area = (np.minimum(b1_x2, b2_x2) - np.maximum(b1_x1, b2_x1)).clip(0) * \
(np.minimum(b1_y2, b2_y2) - np.maximum(b1_y1, b2_y1)).clip(0)
# box2 area
box2_area = (b2_x2 - b2_x1) * (b2_y2 - b2_y1) + 1e-16
# Intersection over box2 area
return inter_area / box2_area
# create random masks
scales = [0.5] * 1 + [0.25] * 2 + [0.125] * 4 + [0.0625] * 8 + [0.03125] * 16 # image size fraction
for s in scales:
mask_h = random.randint(1, int(h * s))
mask_w = random.randint(1, int(w * s))
# box
xmin = max(0, random.randint(0, w) - mask_w // 2)
ymin = max(0, random.randint(0, h) - mask_h // 2)
xmax = min(w, xmin + mask_w)
ymax = min(h, ymin + mask_h)
# apply random color mask
image[ymin:ymax, xmin:xmax] = [random.randint(64, 191) for _ in range(3)]
# return unobscured labels
if len(labels) and s > 0.03:
box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32)
ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area
labels = labels[ioa < 0.60] # remove >60% obscured labels
return labels
def create_folder(path='./new'):
# Create folder
if os.path.exists(path):
shutil.rmtree(path) # delete output folder
os.makedirs(path) # make new output folder
def flatten_recursive(path='../coco128'):
# Flatten a recursive directory by bringing all files to top level
new_path = Path(path + '_flat')
create_folder(new_path)
for file in tqdm(glob.glob(str(Path(path)) + '/**/*.*', recursive=True)):
shutil.copyfile(file, new_path / Path(file).name)
def extract_boxes(path='../coco128/'): # from utils.datasets import *; extract_boxes('../coco128')
# Convert detection dataset into classification dataset, with one directory per class
path = Path(path) # images dir
shutil.rmtree(path / 'classifier') if (path / 'classifier').is_dir() else None # remove existing
files = list(path.rglob('*.*'))
n = len(files) # number of files
for im_file in tqdm(files, total=n):
if im_file.suffix[1:] in img_formats:
# image
im = cv2.imread(str(im_file))[..., ::-1] # BGR to RGB
h, w = im.shape[:2]
# labels
lb_file = Path(img2label_paths([str(im_file)])[0])
if Path(lb_file).exists():
with open(lb_file, 'r') as f:
lb = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32) # labels
for j, x in enumerate(lb):
c = int(x[0]) # class
f = (path / 'classifier') / f'{c}' / f'{path.stem}_{im_file.stem}_{j}.jpg' # new filename
if not f.parent.is_dir():
f.parent.mkdir(parents=True)
b = x[1:] * [w, h, w, h] # box
# b[2:] = b[2:].max() # rectangle to square
b[2:] = b[2:] * 1.2 + 3 # pad
b = xywh2xyxy(b.reshape(-1, 4)).ravel().astype(np.int)
b[[0, 2]] = np.clip(b[[0, 2]], 0, w) # clip boxes outside of image
b[[1, 3]] = np.clip(b[[1, 3]], 0, h)
assert cv2.imwrite(str(f), im[b[1]:b[3], b[0]:b[2]]), f'box failure in {f}'
def autosplit(path='../coco128', weights=(0.9, 0.1, 0.0)): # from utils.datasets import *; autosplit('../coco128')
""" Autosplit a dataset into train/val/test splits and save path/autosplit_*.txt files
# Arguments
path: Path to images directory
weights: Train, val, test weights (list)
"""
path = Path(path) # images dir
files = list(path.rglob('*.*'))
n = len(files) # number of files
indices = random.choices([0, 1, 2], weights=weights, k=n) # assign each image to a split
txt = ['autosplit_train.txt', 'autosplit_val.txt', 'autosplit_test.txt'] # 3 txt files
[(path / x).unlink() for x in txt if (path / x).exists()] # remove existing
for i, img in tqdm(zip(indices, files), total=n):
if img.suffix[1:] in img_formats:
with open(path / txt[i], 'a') as f:
f.write(str(img) + '\n') # add image to txt file
================================================
FILE: utils/general.py
================================================
# General utils
import glob
import logging
import math
import os
import random
import re
import subprocess
import time
from pathlib import Path
import cv2
import numpy as np
import torch
import torchvision
import yaml
from utils.google_utils import gsutil_getsize
from utils.metrics import fitness
from utils.torch_utils import init_torch_seeds
# Settings
torch.set_printoptions(linewidth=320, precision=5, profile='long')
np.set_printoptions(linewidth=320, formatter={'float_kind': '{:11.5g}'.format}) # format short g, %precision=5
cv2.setNumThreads(0) # prevent OpenCV from multithreading (incompatible with PyTorch DataLoader)
os.environ['NUMEXPR_MAX_THREADS'] = str(min(os.cpu_count(), 8)) # NumExpr max threads
def set_logging(rank=-1):
logging.basicConfig(
format="%(message)s",
level=logging.INFO if rank in [-1, 0] else logging.WARN)
def init_seeds(seed=0):
# Initialize random number generator (RNG) seeds
random.seed(seed)
np.random.seed(seed)
init_torch_seeds(seed)
def get_latest_run(search_dir='.'):
# Return path to most recent 'last.pt' in /runs (i.e. to --resume from)
last_list = glob.glob(f'{search_dir}/**/last*.pt', recursive=True)
return max(last_list, key=os.path.getctime) if last_list else ''
def check_online():
# Check internet connectivity
import socket
try:
socket.create_connection(("1.1.1.1", 53)) # check host accesability
return True
except OSError:
return False
def check_git_status():
# Recommend 'git pull' if code is out of date
print(colorstr('github: '), end='')
try:
assert Path('.git').exists(), 'skipping check (not a git repository)'
assert not Path('/workspace').exists(), 'skipping check (Docker image)' # not Path('/.dockerenv').exists()
assert check_online(), 'skipping check (offline)'
cmd = 'git fetch && git config --get remote.origin.url' # github repo url
url = subprocess.check_output(cmd, shell=True).decode()[:-1]
cmd = 'git rev-list $(git rev-parse --abbrev-ref HEAD)..origin/master --count' # commits behind
n = int(subprocess.check_output(cmd, shell=True))
if n > 0:
print(f"⚠️ WARNING: code is out of date by {n} {'commits' if n > 1 else 'commmit'}. "
f"Use 'git pull' to update or 'git clone {url}' to download latest.")
else:
print(f'up to date with {url} ✅')
except Exception as e:
print(e)
def check_requirements(file='requirements.txt'):
# Check installed dependencies meet requirements
import pkg_resources
requirements = pkg_resources.parse_requirements(Path(file).open())
requirements = [x.name + ''.join(*x.specs) if len(x.specs) else x.name for x in requirements]
pkg_resources.require(requirements) # DistributionNotFound or VersionConflict exception if requirements not met
def check_img_size(img_size, s=32):
# Verify img_size is a multiple of stride s
new_size = make_divisible(img_size, int(s)) # ceil gs-multiple
if new_size != img_size:
print('WARNING: --img-size %g must be multiple of max stride %g, updating to %g' % (img_size, s, new_size))
return new_size
def check_file(file):
# Search for file if not found
if os.path.isfile(file) or file == '':
return file
else:
files = glob.glob('./**/' + file, recursive=True) # find file
assert len(files), 'File Not Found: %s' % file # assert file was found
assert len(files) == 1, "Multiple files match '%s', specify exact path: %s" % (file, files) # assert unique
return files[0] # return file
def check_dataset(dict):
# Download dataset if not found locally
val, s = dict.get('val'), dict.get('download')
if val and len(val):
val = [Path(x).resolve() for x in (val if isinstance(val, list) else [val])] # val path
if not all(x.exists() for x in val):
print('\nWARNING: Dataset not found, nonexistent paths: %s' % [str(x) for x in val if not x.exists()])
if s and len(s): # download script
print('Downloading %s ...' % s)
if s.startswith('http') and s.endswith('.zip'): # URL
f = Path(s).name # filename
torch.hub.download_url_to_file(s, f)
r = os.system('unzip -q %s -d ../ && rm %s' % (f, f)) # unzip
else: # bash script
r = os.system(s)
print('Dataset autodownload %s\n' % ('success' if r == 0 else 'failure')) # analyze return value
else:
raise Exception('Dataset not found.')
def make_divisible(x, divisor):
# Returns x evenly divisible by divisor
return math.ceil(x / divisor) * divisor
def clean_str(s):
# Cleans a string by replacing special characters with underscore _
return re.sub(pattern="[|@#!¡·$€%&()=?¿^*;:,¨´><+]", repl="_", string=s)
def one_cycle(y1=0.0, y2=1.0, steps=100):
# lambda function for sinusoidal ramp from y1 to y2
return lambda x: ((1 - math.cos(x * math.pi / steps)) / 2) * (y2 - y1) + y1
def colorstr(*input):
# Colors a string https://en.wikipedia.org/wiki/ANSI_escape_code, i.e. colorstr('blue', 'hello world')
*args, string = input if len(input) > 1 else ('blue', 'bold', input[0]) # color arguments, string
colors = {'black': '\033[30m', # basic colors
'red': '\033[31m',
'green': '\033[32m',
'yellow': '\033[33m',
'blue': '\033[34m',
'magenta': '\0
gitextract_uz5iw4_8/
├── LICENSE
├── README.md
├── face_detector.py
├── models/
│ ├── __init__.py
│ ├── common.py
│ ├── experimental.py
│ ├── export.py
│ ├── yolo.py
│ ├── yolov5-0.5.yaml
│ ├── yolov5l.yaml
│ ├── yolov5l6.yaml
│ ├── yolov5m.yaml
│ ├── yolov5m6.yaml
│ ├── yolov5n.yaml
│ ├── yolov5n6.yaml
│ ├── yolov5s.yaml
│ └── yolov5s6.yaml
├── requirements.txt
├── utils/
│ ├── __init__.py
│ ├── activations.py
│ ├── autoanchor.py
│ ├── aws/
│ │ ├── __init__.py
│ │ ├── mime.sh
│ │ ├── resume.py
│ │ └── userdata.sh
│ ├── datasets.py
│ ├── face_datasets.py
│ ├── general.py
│ ├── google_app_engine/
│ │ ├── Dockerfile
│ │ ├── additional_requirements.txt
│ │ └── app.yaml
│ ├── google_utils.py
│ ├── infer_utils.py
│ ├── loss.py
│ ├── metrics.py
│ ├── plots.py
│ ├── preprocess_utils.py
│ ├── torch_utils.py
│ └── wandb_logging/
│ ├── __init__.py
│ ├── log_dataset.py
│ └── wandb_utils.py
└── weights/
├── download_weights.sh
└── yolov5n_state_dict.pt
SYMBOL INDEX (325 symbols across 18 files)
FILE: face_detector.py
class YoloDetector (line 22) | class YoloDetector:
method __init__ (line 23) | def __init__(self, weights_name='yolov5n_state_dict.pt', config_name='...
method init_detector (line 42) | def init_detector(self,weights_name,config_name):
method _preprocess (line 58) | def _preprocess(self,imgs):
method _postprocess (line 80) | def _postprocess(self, imgs, origimgs, pred, conf_thres, iou_thres):
method get_frontal_predict (line 113) | def get_frontal_predict(self, box, points):
method align (line 130) | def align(self, img, points):
method predict (line 142) | def predict(self, imgs, conf_thres = 0.3, iou_thres = 0.5):
method __call__ (line 187) | def __call__(self,*args):
FILE: models/common.py
function autopad (line 15) | def autopad(k, p=None): # kernel, padding
function channel_shuffle (line 21) | def channel_shuffle(x, groups):
function DWConv (line 33) | def DWConv(c1, c2, k=1, s=1, act=True):
class Conv (line 37) | class Conv(nn.Module):
method __init__ (line 39) | def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in,...
method forward (line 46) | def forward(self, x):
method fuseforward (line 49) | def fuseforward(self, x):
class StemBlock (line 52) | class StemBlock(nn.Module):
method __init__ (line 53) | def __init__(self, c1, c2, k=3, s=2, p=None, g=1, act=True):
method forward (line 61) | def forward(self, x):
class Bottleneck (line 69) | class Bottleneck(nn.Module):
method __init__ (line 71) | def __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_ou...
method forward (line 78) | def forward(self, x):
class BottleneckCSP (line 81) | class BottleneckCSP(nn.Module):
method __init__ (line 83) | def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ...
method forward (line 94) | def forward(self, x):
class C3 (line 100) | class C3(nn.Module):
method __init__ (line 102) | def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ...
method forward (line 110) | def forward(self, x):
class ShuffleV2Block (line 113) | class ShuffleV2Block(nn.Module):
method __init__ (line 114) | def __init__(self, inp, oup, stride):
method depthwise_conv (line 147) | def depthwise_conv(i, o, kernel_size, stride=1, padding=0, bias=False):
method forward (line 150) | def forward(self, x):
class SPP (line 159) | class SPP(nn.Module):
method __init__ (line 161) | def __init__(self, c1, c2, k=(5, 9, 13)):
method forward (line 168) | def forward(self, x):
class Focus (line 173) | class Focus(nn.Module):
method __init__ (line 175) | def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in,...
method forward (line 180) | def forward(self, x): # x(b,c,w,h) -> y(b,4c,w/2,h/2)
class Contract (line 185) | class Contract(nn.Module):
method __init__ (line 187) | def __init__(self, gain=2):
method forward (line 191) | def forward(self, x):
class Expand (line 199) | class Expand(nn.Module):
method __init__ (line 201) | def __init__(self, gain=2):
method forward (line 205) | def forward(self, x):
class Concat (line 213) | class Concat(nn.Module):
method __init__ (line 215) | def __init__(self, dimension=1):
method forward (line 219) | def forward(self, x):
class NMS (line 223) | class NMS(nn.Module):
method __init__ (line 229) | def __init__(self):
method forward (line 232) | def forward(self, x):
class autoShape (line 235) | class autoShape(nn.Module):
method __init__ (line 242) | def __init__(self, model):
method autoshape (line 246) | def autoshape(self):
method forward (line 250) | def forward(self, imgs, size=640, augment=False, profile=False):
class Detections (line 297) | class Detections:
method __init__ (line 299) | def __init__(self, imgs, pred, names=None):
method display (line 312) | def display(self, pprint=False, show=False, save=False, render=False):
method print (line 336) | def print(self):
method show (line 339) | def show(self):
method save (line 342) | def save(self):
method render (line 345) | def render(self):
method __len__ (line 349) | def __len__(self):
method tolist (line 352) | def tolist(self):
class Classify (line 361) | class Classify(nn.Module):
method __init__ (line 363) | def __init__(self, c1, c2, k=1, s=1, p=None, g=1): # ch_in, ch_out, k...
method forward (line 369) | def forward(self, x):
FILE: models/experimental.py
class CrossConv (line 11) | class CrossConv(nn.Module):
method __init__ (line 13) | def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False):
method forward (line 21) | def forward(self, x):
class Sum (line 25) | class Sum(nn.Module):
method __init__ (line 27) | def __init__(self, n, weight=False): # n: number of inputs
method forward (line 34) | def forward(self, x):
class GhostConv (line 46) | class GhostConv(nn.Module):
method __init__ (line 48) | def __init__(self, c1, c2, k=1, s=1, g=1, act=True): # ch_in, ch_out,...
method forward (line 54) | def forward(self, x):
class GhostBottleneck (line 59) | class GhostBottleneck(nn.Module):
method __init__ (line 61) | def __init__(self, c1, c2, k, s):
method forward (line 70) | def forward(self, x):
class MixConv2d (line 74) | class MixConv2d(nn.Module):
method __init__ (line 76) | def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):
method forward (line 94) | def forward(self, x):
class Ensemble (line 98) | class Ensemble(nn.ModuleList):
method __init__ (line 100) | def __init__(self):
method forward (line 103) | def forward(self, x, augment=False):
function attempt_load (line 113) | def attempt_load(weights, map_location=None):
FILE: models/yolo.py
class Detect (line 27) | class Detect(nn.Module):
method __init__ (line 31) | def __init__(self, nc=80, anchors=(), ch=()): # detection layer
method forward (line 45) | def forward(self, x):
method _make_grid (line 88) | def _make_grid(nx=20, ny=20):
class Model (line 93) | class Model(nn.Module):
method __init__ (line 94) | def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None): # model, input...
method forward (line 129) | def forward(self, x, augment=False, profile=False):
method forward_once (line 149) | def forward_once(self, x, profile=False):
method _initialize_biases (line 170) | def _initialize_biases(self, cf=None): # initialize biases into Detec...
method _print_biases (line 180) | def _print_biases(self):
method fuse (line 191) | def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers
method nms (line 201) | def nms(self, mode=True): # add or remove NMS module
method autoshape (line 215) | def autoshape(self): # add autoShape module
method info (line 221) | def info(self, verbose=False, img_size=640): # print model information
function parse_model (line 225) | def parse_model(d, ch): # model_dict, input_channels(3)
FILE: utils/activations.py
class SiLU (line 9) | class SiLU(nn.Module): # export-friendly version of nn.SiLU()
method forward (line 11) | def forward(x):
class Hardswish (line 15) | class Hardswish(nn.Module): # export-friendly version of nn.Hardswish()
method forward (line 17) | def forward(x):
class MemoryEfficientSwish (line 22) | class MemoryEfficientSwish(nn.Module):
class F (line 23) | class F(torch.autograd.Function):
method forward (line 25) | def forward(ctx, x):
method backward (line 30) | def backward(ctx, grad_output):
method forward (line 35) | def forward(self, x):
class Mish (line 40) | class Mish(nn.Module):
method forward (line 42) | def forward(x):
class MemoryEfficientMish (line 46) | class MemoryEfficientMish(nn.Module):
class F (line 47) | class F(torch.autograd.Function):
method forward (line 49) | def forward(ctx, x):
method backward (line 54) | def backward(ctx, grad_output):
method forward (line 60) | def forward(self, x):
class FReLU (line 65) | class FReLU(nn.Module):
method __init__ (line 66) | def __init__(self, c1, k=3): # ch_in, kernel
method forward (line 71) | def forward(self, x):
FILE: utils/autoanchor.py
function check_anchor_order (line 12) | def check_anchor_order(m):
function check_anchors (line 23) | def check_anchors(dataset, model, thr=4.0, imgsz=640):
function kmean_anchors (line 58) | def kmean_anchors(path='./data/coco128.yaml', n=9, img_size=640, thr=4.0...
FILE: utils/datasets.py
function get_hash (line 38) | def get_hash(files):
function exif_size (line 43) | def exif_size(img):
function create_dataloader (line 58) | def create_dataloader(path, imgsz, batch_size, stride, opt, hyp=None, au...
class InfiniteDataLoader (line 87) | class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):
method __init__ (line 93) | def __init__(self, *args, **kwargs):
method __len__ (line 98) | def __len__(self):
method __iter__ (line 101) | def __iter__(self):
class _RepeatSampler (line 106) | class _RepeatSampler(object):
method __init__ (line 113) | def __init__(self, sampler):
method __iter__ (line 116) | def __iter__(self):
class LoadImages (line 121) | class LoadImages: # for inference
method __init__ (line 122) | def __init__(self, path, img_size=640):
method __iter__ (line 150) | def __iter__(self):
method __next__ (line 154) | def __next__(self):
method new_video (line 192) | def new_video(self, path):
method __len__ (line 197) | def __len__(self):
class LoadWebcam (line 201) | class LoadWebcam: # for inference
method __init__ (line 202) | def __init__(self, pipe='0', img_size=640):
method __iter__ (line 215) | def __iter__(self):
method __next__ (line 219) | def __next__(self):
method __len__ (line 254) | def __len__(self):
class LoadStreams (line 258) | class LoadStreams: # multiple IP or RTSP cameras
method __init__ (line 259) | def __init__(self, sources='streams.txt', img_size=640):
method update (line 292) | def update(self, index, cap):
method __iter__ (line 304) | def __iter__(self):
method __next__ (line 308) | def __next__(self):
method __len__ (line 327) | def __len__(self):
function img2label_paths (line 331) | def img2label_paths(img_paths):
class LoadImagesAndLabels (line 337) | class LoadImagesAndLabels(Dataset): # for training/testing
method __init__ (line 338) | def __init__(self, path, img_size=640, batch_size=16, augment=False, h...
method cache_labels (line 437) | def cache_labels(self, path=Path('./labels.cache'), prefix=''):
method __len__ (line 483) | def __len__(self):
method __getitem__ (line 492) | def __getitem__(self, index):
method collate_fn (line 569) | def collate_fn(batch):
method collate_fn4 (line 576) | def collate_fn4(batch):
function load_image (line 603) | def load_image(self, index):
function augment_hsv (line 620) | def augment_hsv(img, hgain=0.5, sgain=0.5, vgain=0.5):
function load_mosaic (line 639) | def load_mosaic(self, index):
function load_mosaic9 (line 693) | def load_mosaic9(self, index):
function replicate (line 763) | def replicate(img, labels):
function letterbox (line 780) | def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=Tru...
function random_perspective (line 813) | def random_perspective(img, targets=(), degrees=10, translate=.1, scale=...
function box_candidates (line 900) | def box_candidates(box1, box2, wh_thr=2, ar_thr=20, area_thr=0.1, eps=1e...
function cutout (line 908) | def cutout(image, labels):
function create_folder (line 954) | def create_folder(path='./new'):
function flatten_recursive (line 961) | def flatten_recursive(path='../coco128'):
function extract_boxes (line 969) | def extract_boxes(path='../coco128/'): # from utils.datasets import *; ...
function autosplit (line 1004) | def autosplit(path='../coco128', weights=(0.9, 0.1, 0.0)): # from utils...
FILE: utils/face_datasets.py
function get_hash (line 35) | def get_hash(files):
function img2label_paths (line 39) | def img2label_paths(img_paths):
function exif_size (line 44) | def exif_size(img):
function create_dataloader (line 58) | def create_dataloader(path, imgsz, batch_size, stride, opt, hyp=None, au...
class InfiniteDataLoader (line 85) | class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):
method __init__ (line 91) | def __init__(self, *args, **kwargs):
method __len__ (line 96) | def __len__(self):
method __iter__ (line 99) | def __iter__(self):
class _RepeatSampler (line 102) | class _RepeatSampler(object):
method __init__ (line 109) | def __init__(self, sampler):
method __iter__ (line 112) | def __iter__(self):
class LoadFaceImagesAndLabels (line 116) | class LoadFaceImagesAndLabels(Dataset): # for training/testing
method __init__ (line 117) | def __init__(self, path, img_size=640, batch_size=16, augment=False, h...
method cache_labels (line 216) | def cache_labels(self, path=Path('./labels.cache')):
method __len__ (line 262) | def __len__(self):
method __getitem__ (line 271) | def __getitem__(self, index):
method collate_fn (line 405) | def collate_fn(batch):
function showlabels (line 412) | def showlabels(img, boxs, landmarks):
function load_mosaic_face (line 426) | def load_mosaic_face(self, index):
function load_image (line 515) | def load_image(self, index):
function augment_hsv (line 532) | def augment_hsv(img, hgain=0.5, sgain=0.5, vgain=0.5):
function replicate (line 550) | def replicate(img, labels):
function letterbox (line 567) | def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=Tru...
function random_perspective (line 600) | def random_perspective(img, targets=(), degrees=10, translate=.1, scale=...
function box_candidates (line 715) | def box_candidates(box1, box2, wh_thr=2, ar_thr=20, area_thr=0.1): # bo...
function cutout (line 723) | def cutout(image, labels):
function create_folder (line 769) | def create_folder(path='./new'):
function flatten_recursive (line 776) | def flatten_recursive(path='../coco128'):
function extract_boxes (line 784) | def extract_boxes(path='../coco128/'): # from utils.datasets import *; ...
function autosplit (line 819) | def autosplit(path='../coco128', weights=(0.9, 0.1, 0.0)): # from utils...
FILE: utils/general.py
function set_logging (line 30) | def set_logging(rank=-1):
function init_seeds (line 36) | def init_seeds(seed=0):
function get_latest_run (line 43) | def get_latest_run(search_dir='.'):
function check_online (line 49) | def check_online():
function check_git_status (line 59) | def check_git_status():
function check_requirements (line 80) | def check_requirements(file='requirements.txt'):
function check_img_size (line 88) | def check_img_size(img_size, s=32):
function check_file (line 96) | def check_file(file):
function check_dataset (line 107) | def check_dataset(dict):
function make_divisible (line 127) | def make_divisible(x, divisor):
function clean_str (line 132) | def clean_str(s):
function one_cycle (line 137) | def one_cycle(y1=0.0, y2=1.0, steps=100):
function colorstr (line 142) | def colorstr(*input):
function labels_to_class_weights (line 167) | def labels_to_class_weights(labels, nc=80):
function labels_to_image_weights (line 186) | def labels_to_image_weights(labels, nc=80, class_weights=np.ones(80)):
function coco80_to_coco91_class (line 194) | def coco80_to_coco91_class(): # converts 80-index (val2014) to 91-index...
function xyxy2xywh (line 206) | def xyxy2xywh(x):
function xywh2xyxy (line 216) | def xywh2xyxy(x):
function xywhn2xyxy (line 226) | def xywhn2xyxy(x, w=640, h=640, padw=32, padh=32):
function scale_coords (line 236) | def scale_coords(img1_shape, coords, img0_shape, ratio_pad=None):
function clip_coords (line 252) | def clip_coords(boxes, img_shape):
function bbox_iou (line 260) | def bbox_iou(box1, box2, x1y1x2y2=True, GIoU=False, DIoU=False, CIoU=Fal...
function box_iou (line 307) | def box_iou(box1, box2):
function wh_iou (line 334) | def wh_iou(wh1, wh2):
function jaccard_diou (line 342) | def jaccard_diou(box_a, box_b, iscrowd:bool=False):
function non_max_suppression_face (line 379) | def non_max_suppression_face(prediction, conf_thres=0.25, iou_thres=0.45...
function non_max_suppression (line 459) | def non_max_suppression(prediction, conf_thres=0.25, iou_thres=0.45, cla...
function strip_optimizer (line 552) | def strip_optimizer(f='weights/best.pt', s=''): # from utils.general im...
function print_mutation (line 566) | def print_mutation(hyp, results, yaml_file='hyp_evolved.yaml', bucket=''):
function apply_classifier (line 597) | def apply_classifier(x, model, img, im0):
function increment_path (line 636) | def increment_path(path, exist_ok=True, sep=''):
function filter_boxes (line 647) | def filter_boxes(boxes, min_size):
function scale_coords_landmarks (line 654) | def scale_coords_landmarks(img1_shape, coords, img0_shape, ratio_pad=None):
FILE: utils/google_utils.py
function gsutil_getsize (line 13) | def gsutil_getsize(url=''):
function attempt_download (line 19) | def attempt_download(file, repo='ultralytics/yolov5'):
function gdrive_download (line 55) | def gdrive_download(id='16TiPfZj7htmTyhntwcZyEEAejOUxuT6m', file='tmp.zi...
function get_token (line 90) | def get_token(cookie="./cookie"):
FILE: utils/infer_utils.py
function decode_infer (line 5) | def decode_infer(output, stride):
FILE: utils/loss.py
function smooth_BCE (line 10) | def smooth_BCE(eps=0.1): # https://github.com/ultralytics/yolov3/issues...
class BCEBlurWithLogitsLoss (line 15) | class BCEBlurWithLogitsLoss(nn.Module):
method __init__ (line 17) | def __init__(self, alpha=0.05):
method forward (line 22) | def forward(self, pred, true):
class FocalLoss (line 32) | class FocalLoss(nn.Module):
method __init__ (line 34) | def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
method forward (line 42) | def forward(self, pred, true):
class QFocalLoss (line 62) | class QFocalLoss(nn.Module):
method __init__ (line 64) | def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
method forward (line 72) | def forward(self, pred, true):
class WingLoss (line 87) | class WingLoss(nn.Module):
method __init__ (line 88) | def __init__(self, w=10, e=2):
method forward (line 95) | def forward(self, x, t, sigma=1):
class LandmarksLoss (line 104) | class LandmarksLoss(nn.Module):
method __init__ (line 106) | def __init__(self, alpha=1.0):
method forward (line 111) | def forward(self, pred, truel, mask):
function compute_loss (line 116) | def compute_loss(p, targets, model): # predictions, targets, model
function build_targets (line 196) | def build_targets(p, targets, model):
FILE: utils/metrics.py
function fitness (line 12) | def fitness(x):
function ap_per_class (line 18) | def ap_per_class(tp, conf, pred_cls, target_cls, plot=False, save_dir='p...
function compute_ap (line 79) | def compute_ap(recall, precision):
class ConfusionMatrix (line 107) | class ConfusionMatrix:
method __init__ (line 109) | def __init__(self, nc, conf=0.25, iou_thres=0.45):
method process_batch (line 115) | def process_batch(self, detections, labels):
method matrix (line 155) | def matrix(self):
method plot (line 158) | def plot(self, save_dir='', names=()):
method print (line 177) | def print(self):
function plot_pr_curve (line 184) | def plot_pr_curve(px, py, ap, save_dir='.', names=()):
FILE: utils/plots.py
function color_list (line 29) | def color_list():
function hist2d (line 37) | def hist2d(x, y, n=100):
function butter_lowpass_filtfilt (line 46) | def butter_lowpass_filtfilt(data, cutoff=1500, fs=50000, order=5):
function plot_one_box (line 57) | def plot_one_box(x, img, color=None, label=None, line_thickness=None):
function plot_wh_methods (line 71) | def plot_wh_methods(): # from utils.plots import *; plot_wh_methods()
function output_to_target (line 91) | def output_to_target(output):
function plot_images (line 100) | def plot_images(images, targets, paths=None, fname='images.jpg', names=N...
function plot_lr_scheduler (line 179) | def plot_lr_scheduler(optimizer, scheduler, epochs=300, save_dir=''):
function plot_test_txt (line 196) | def plot_test_txt(): # from utils.plots import *; plot_test()
function plot_targets_txt (line 213) | def plot_targets_txt(): # from utils.plots import *; plot_targets_txt()
function plot_study_txt (line 226) | def plot_study_txt(path='study/', x=None): # from utils.plots import *;...
function plot_labels (line 257) | def plot_labels(labels, save_dir=Path(''), loggers=None):
function plot_evolution (line 301) | def plot_evolution(yaml_file='data/hyp.finetune.yaml'): # from utils.pl...
function profile_idetection (line 325) | def profile_idetection(start=0, stop=0, labels=(), save_dir=''):
function plot_results_overlay (line 357) | def plot_results_overlay(start=0, stop=0): # from utils.plots import *;...
function plot_results (line 380) | def plot_results(start=0, stop=0, bucket='', id=(), labels=(), save_dir=...
FILE: utils/preprocess_utils.py
function align_faces (line 3) | def align_faces(img, bbox=None, landmark=None, **kwargs):
function face_distance (line 25) | def face_distance(vec1,vec2):
FILE: utils/torch_utils.py
function torch_distributed_zero_first (line 26) | def torch_distributed_zero_first(local_rank: int):
function init_torch_seeds (line 37) | def init_torch_seeds(seed=0):
function git_describe (line 46) | def git_describe():
function select_device (line 54) | def select_device(device='', batch_size=None):
function time_synchronized (line 80) | def time_synchronized():
function profile (line 87) | def profile(x, ops, n=100, device=None):
function is_parallel (line 126) | def is_parallel(model):
function intersect_dicts (line 130) | def intersect_dicts(da, db, exclude=()):
function initialize_weights (line 135) | def initialize_weights(model):
function find_modules (line 147) | def find_modules(model, mclass=nn.Conv2d):
function sparsity (line 152) | def sparsity(model):
function prune (line 161) | def prune(model, amount=0.3):
function fuse_conv_and_bn (line 172) | def fuse_conv_and_bn(conv, bn):
function model_info (line 195) | def model_info(model, verbose=False, img_size=640):
function load_classifier (line 219) | def load_classifier(name='resnet101', n=2):
function scale_img (line 238) | def scale_img(img, ratio=1.0, same_shape=False, gs=32): # img(16,3,256,...
function copy_attr (line 251) | def copy_attr(a, b, include=(), exclude=()):
class ModelEMA (line 260) | class ModelEMA:
method __init__ (line 270) | def __init__(self, model, decay=0.9999, updates=0):
method update (line 280) | def update(self, model):
method update_attr (line 292) | def update_attr(self, model, include=(), exclude=('process_group', 're...
FILE: utils/wandb_logging/log_dataset.py
function create_dataset_artifact (line 10) | def create_dataset_artifact(opt):
FILE: utils/wandb_logging/wandb_utils.py
function remove_prefix (line 23) | def remove_prefix(from_string, prefix=WANDB_ARTIFACT_PREFIX):
function check_wandb_config_file (line 27) | def check_wandb_config_file(data_config_file):
function get_run_info (line 34) | def get_run_info(run_path):
function check_wandb_resume (line 42) | def check_wandb_resume(opt):
function process_wandb_config_ddp_mode (line 56) | def process_wandb_config_ddp_mode(opt):
class WandbLogger (line 80) | class WandbLogger():
method __init__ (line 81) | def __init__(self, opt, name, run_id, data_dict, job_type='Training'):
method check_and_upload_dataset (line 115) | def check_and_upload_dataset(self, opt):
method setup_training (line 126) | def setup_training(self, opt, data_dict):
method download_dataset_artifact (line 159) | def download_dataset_artifact(self, path, alias):
method download_model_artifact (line 167) | def download_model_artifact(self, opt):
method log_model (line 179) | def log_model(self, path, opt, epoch, fitness_score, best_model=False):
method log_dataset_artifact (line 193) | def log_dataset_artifact(self, data_file, single_cls, project, overwri...
method map_val_table_path (line 222) | def map_val_table_path(self):
method create_dataset_table (line 228) | def create_dataset_table(self, dataset, class_to_id, name='dataset'):
method log_training_progress (line 263) | def log_training_progress(self, predn, path, names):
method log (line 285) | def log(self, log_dict):
method end_epoch (line 290) | def end_epoch(self, best_result=False):
method finish_run (line 302) | def finish_run(self):
Condensed preview — 43 files, each showing path, character count, and a content snippet. Download the .json file or copy for the full structured content (305K chars).
[
{
"path": "LICENSE",
"chars": 35127,
"preview": "GNU GENERAL PUBLIC LICENSE\n Version 3, 29 June 2007\n\n Copyright (C) 2007 Free Software Foundation,"
},
{
"path": "README.md",
"chars": 2313,
"preview": "# Yolov5 Face Detection\n\n## Description\nThe project is a wrap over [yolov5-face](https://github.com/deepcam-cn/yolov5-fa"
},
{
"path": "face_detector.py",
"chars": 8592,
"preview": "import joblib\nimport os\nimport sys\nimport torch\nimport torch.nn as nn\nimport numpy as np\nimport cv2\nimport copy\nimport s"
},
{
"path": "models/__init__.py",
"chars": 0,
"preview": ""
},
{
"path": "models/common.py",
"chars": 15485,
"preview": "# This file contains modules common to various models\n\nimport math\n\nimport numpy as np\nimport requests\nimport torch\nimpo"
},
{
"path": "models/experimental.py",
"chars": 5056,
"preview": "# This file contains experimental modules\n\nimport numpy as np\nimport torch\nimport torch.nn as nn\n\nfrom models.common imp"
},
{
"path": "models/export.py",
"chars": 2855,
"preview": "\"\"\"Exports a YOLOv5 *.pt model to ONNX and TorchScript formats\n\nUsage:\n $ export PYTHONPATH=\"$PWD\" && python models/e"
},
{
"path": "models/yolo.py",
"chars": 13997,
"preview": "import argparse\nimport logging\nimport math\nimport sys\nfrom copy import deepcopy\nfrom pathlib import Path\n\nimport torch\ni"
},
{
"path": "models/yolov5-0.5.yaml",
"chars": 1346,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 1.0 # model depth multiple\nwidth_multiple: 0.5 # layer channel"
},
{
"path": "models/yolov5l.yaml",
"chars": 1345,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 1.0 # model depth multiple\nwidth_multiple: 1.0 # layer channel"
},
{
"path": "models/yolov5l6.yaml",
"chars": 1921,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 1.0 # model depth multiple\nwidth_multiple: 1.0 # layer channel"
},
{
"path": "models/yolov5m.yaml",
"chars": 1347,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 0.67 # model depth multiple\nwidth_multiple: 0.75 # layer chann"
},
{
"path": "models/yolov5m6.yaml",
"chars": 1923,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 0.67 # model depth multiple\nwidth_multiple: 0.75 # layer chann"
},
{
"path": "models/yolov5n.yaml",
"chars": 1346,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 1.0 # model depth multiple\nwidth_multiple: 1.0 # layer channel"
},
{
"path": "models/yolov5n6.yaml",
"chars": 1784,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 1.0 # model depth multiple\nwidth_multiple: 1.0 # layer channel"
},
{
"path": "models/yolov5s.yaml",
"chars": 1347,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 0.33 # model depth multiple\nwidth_multiple: 0.35 # layer chann"
},
{
"path": "models/yolov5s6.yaml",
"chars": 1923,
"preview": "# parameters\nnc: 1 # number of classes\ndepth_multiple: 0.33 # model depth multiple\nwidth_multiple: 0.50 # layer chann"
},
{
"path": "requirements.txt",
"chars": 288,
"preview": "joblib==1.2.0\nmatplotlib==3.5.1\nnumpy==1.22.4\nonnx==1.12.0\nopencv_python==4.6.0.66\npandas==1.4.2\nPillow==9.3.0\nPyYAML==6"
},
{
"path": "utils/__init__.py",
"chars": 0,
"preview": ""
},
{
"path": "utils/activations.py",
"chars": 2248,
"preview": "# Activation functions\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\n\n# SiLU https://arxiv.org/pd"
},
{
"path": "utils/autoanchor.py",
"chars": 6940,
"preview": "# Auto-anchor utils\n\nimport numpy as np\nimport torch\nimport yaml\nfrom scipy.cluster.vq import kmeans\nfrom tqdm import tq"
},
{
"path": "utils/aws/__init__.py",
"chars": 0,
"preview": ""
},
{
"path": "utils/aws/mime.sh",
"chars": 780,
"preview": "# AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/\n#"
},
{
"path": "utils/aws/resume.py",
"chars": 1114,
"preview": "# Resume all interrupted trainings in yolov5/ dir including DDP trainings\n# Usage: $ python utils/aws/resume.py\n\nimport "
},
{
"path": "utils/aws/userdata.sh",
"chars": 1237,
"preview": "#!/bin/bash\n# AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html\n# This "
},
{
"path": "utils/datasets.py",
"chars": 41615,
"preview": "# Dataset utils and dataloaders\n\nimport glob\nimport logging\nimport math\nimport os\nimport random\nimport shutil\nimport tim"
},
{
"path": "utils/face_datasets.py",
"chars": 38959,
"preview": "import glob\nimport logging\nimport math\nimport os\nimport random\nimport shutil\nimport time\nfrom itertools import repeat\nfr"
},
{
"path": "utils/general.py",
"chars": 28226,
"preview": "# General utils\n\nimport glob\nimport logging\nimport math\nimport os\nimport random\nimport re\nimport subprocess\nimport time\n"
},
{
"path": "utils/google_app_engine/Dockerfile",
"chars": 821,
"preview": "FROM gcr.io/google-appengine/python\n\n# Create a virtualenv for dependencies. This isolates these packages from\n# system-"
},
{
"path": "utils/google_app_engine/additional_requirements.txt",
"chars": 105,
"preview": "# add these requirements in your app on top of the existing ones\npip==18.1\nFlask==1.0.2\ngunicorn==19.9.0\n"
},
{
"path": "utils/google_app_engine/app.yaml",
"chars": 173,
"preview": "runtime: custom\nenv: flex\n\nservice: yolov5app\n\nliveness_check:\n initial_delay_sec: 600\n\nmanual_scaling:\n instances: 1\n"
},
{
"path": "utils/google_utils.py",
"chars": 4879,
"preview": "# Google utils: https://cloud.google.com/storage/docs/reference/libraries\n\nimport os\nimport platform\nimport subprocess\ni"
},
{
"path": "utils/infer_utils.py",
"chars": 1281,
"preview": "import torch\n\n\n\ndef decode_infer(output, stride):\n # logging.info(torch.tensor(output.shape[0]))\n # logging.info(o"
},
{
"path": "utils/loss.py",
"chars": 13155,
"preview": "# Loss functions\n\nimport torch\nimport torch.nn as nn\nimport numpy as np\nfrom utils.general import bbox_iou\nfrom utils.to"
},
{
"path": "utils/metrics.py",
"chars": 7950,
"preview": "# Model validation metrics\n\nfrom pathlib import Path\n\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport torch\n\nf"
},
{
"path": "utils/plots.py",
"chars": 17309,
"preview": "# Plotting utils\n\nimport glob\nimport math\nimport os\nimport random\nfrom copy import copy\nfrom pathlib import Path\n\nimport"
},
{
"path": "utils/preprocess_utils.py",
"chars": 890,
"preview": "import numpy as np\nimport cv2\ndef align_faces(img, bbox=None, landmark=None, **kwargs):\n M = None\n # Do alignment "
},
{
"path": "utils/torch_utils.py",
"chars": 11947,
"preview": "# PyTorch utils\n\nimport logging\nimport math\nimport os\nimport subprocess\nimport time\nfrom contextlib import contextmanage"
},
{
"path": "utils/wandb_logging/__init__.py",
"chars": 0,
"preview": ""
},
{
"path": "utils/wandb_logging/log_dataset.py",
"chars": 819,
"preview": "import argparse\n\nimport yaml\n\nfrom wandb_utils import WandbLogger\n\nWANDB_ARTIFACT_PREFIX = 'wandb-artifact://'\n\n\ndef cre"
},
{
"path": "utils/wandb_logging/wandb_utils.py",
"chars": 16268,
"preview": "import json\nimport sys\nfrom pathlib import Path\n\nimport torch\nimport yaml\nfrom tqdm import tqdm\n\nsys.path.append(str(Pat"
},
{
"path": "weights/download_weights.sh",
"chars": 278,
"preview": "#!/bin/bash\n# Download latest models from https://github.com/ultralytics/yolov5/releases\n# Usage:\n# $ bash weights/do"
}
]
// ... and 1 more files (download for full content)
About this extraction
This page contains the full source code of the elyha7/yoloface GitHub repository, extracted and formatted as plain text for AI agents and large language models (LLMs). The extraction includes 43 files (288.1 KB), approximately 88.8k tokens, and a symbol index with 325 extracted functions, classes, methods, constants, and types. Use this with OpenClaw, Claude, ChatGPT, Cursor, Windsurf, or any other AI tool that accepts text input. You can copy the full output to your clipboard or download it as a .txt file.
Extracted by GitExtract — free GitHub repo to text converter for AI. Built by Nikandr Surkov.