labelme
Image annotation with Python.
VOC dataset example of instance segmentation.
Other examples (semantic segmentation, bbox detection, and classification).
Various primitives (polygon, rectangle, circle, line, and point).
Multi-language support (English, 中文, 日本語, 한국어, Deutsch, Français, and more).
## Features - [x] Image annotation for polygon, rectangle, circle, line and point ([tutorial](examples/tutorial)) - [x] Image flag annotation for classification and cleaning ([#166](https://github.com/wkentaro/labelme/pull/166)) - [x] Video annotation ([video annotation](examples/video_annotation)) - [x] GUI customization (predefined labels / flags, auto-saving, label validation, etc) ([#144](https://github.com/wkentaro/labelme/pull/144)) - [x] Exporting VOC-format dataset for [semantic segmentation](examples/semantic_segmentation), [instance segmentation](examples/instance_segmentation) - [x] Exporting COCO-format dataset for [instance segmentation](examples/instance_segmentation) - [x] AI-assisted point-to-polygon/mask annotation by SAM, EfficientSAM models - [x] AI text-to-annotation by YOLO-world, SAM3 models **🌏 Available in 16 languages** - English · 日本語 · 한국어 · 简体中文 · 繁體中文 · Deutsch · Français · Español · Italiano · Português · Nederlands · Magyar · Tiếng Việt · Türkçe · Polski · فارسی (`LANG=ja_JP.UTF-8 labelme`) ## Installation There are 3 options to install labelme: ### Option 1: Using pip For more detail, check ["Install Labelme using Terminal"](https://www.labelme.io/docs/install-labelme-terminal) ```bash pip install labelme # To install the latest version from GitHub: # pip install git+https://github.com/wkentaro/labelme.git ``` ### Option 2: Using standalone executable (Easiest) If you're willing to invest in the convenience of simple installation without any dependencies (Python, Qt), you can download the standalone executable from ["Install Labelme as App"](https://www.labelme.io/docs/install-labelme-app). It's a one-time payment for lifetime access, and it helps us to maintain this project. ### Option 3: Using a package manager in each Linux distribution In some Linux distributions, you can install labelme via their package managers (e.g., apt, pacman). The following systems are currently available: [](https://repology.org/project/labelme/versions) ## Usage Run `labelme --help` for detail. The annotations are saved as a [JSON](http://www.json.org/) file. ```bash labelme # just open gui # tutorial (single image example) cd examples/tutorial labelme apc2016_obj3.jpg # specify image file labelme apc2016_obj3.jpg --output annotations/ # save annotation JSON files to a directory labelme apc2016_obj3.jpg --with-image-data # include image data in JSON file labelme apc2016_obj3.jpg \ --labels highland_6539_self_stick_notes,mead_index_cards,kong_air_dog_squeakair_tennis_ball # specify label list # semantic segmentation example cd examples/semantic_segmentation labelme data_annotated/ # Open directory to annotate all images in it labelme data_annotated/ --labels labels.txt # specify label list with a file ``` ### Command Line Arguments - `--output` specifies the location that annotations will be written to. If the location ends with .json, a single annotation will be written to this file. Only one image can be annotated if a location is specified with .json. If the location does not end with .json, the program will assume it is a directory. Annotations will be stored in this directory with a name that corresponds to the image that the annotation was made on. - The first time you run labelme, it will create a config file at `~/.labelmerc`. Add only the settings you want to override. For all available options and their defaults, see [`default_config.yaml`](labelme/config/default_config.yaml). If you would prefer to use a config file from another location, you can specify this file with the `--config` flag. - Without the `--nosortlabels` flag, the program will list labels in alphabetical order. When the program is run with this flag, it will display labels in the order that they are provided. - Flags are assigned to an entire image. [Example](examples/classification) - Labels are assigned to a single polygon. [Example](examples/bbox_detection) ### FAQ - **How to convert JSON file to numpy array?** See [examples/tutorial](examples/tutorial#convert-to-dataset). - **How to load label PNG file?** See [examples/tutorial](examples/tutorial#how-to-load-label-png-file). - **How to get annotations for semantic segmentation?** See [examples/semantic_segmentation](examples/semantic_segmentation). - **How to get annotations for instance segmentation?** See [examples/instance_segmentation](examples/instance_segmentation). ## Examples * [Image Classification](examples/classification) * [Bounding Box Detection](examples/bbox_detection) * [Semantic Segmentation](examples/semantic_segmentation) * [Instance Segmentation](examples/instance_segmentation) * [Video Annotation](examples/video_annotation) ## How to build standalone executable ```bash LABELME_PATH=./labelme OSAM_PATH=$(python -c 'import os, osam; print(os.path.dirname(osam.__file__))') pip install 'numpy<2.0' # numpy>=2.0 causes build errors (see #1532) pyinstaller labelme/labelme/__main__.py \ --name=Labelme \ --windowed \ --noconfirm \ --specpath=build \ --add-data=$(OSAM_PATH)/_models/yoloworld/clip/bpe_simple_vocab_16e6.txt.gz:osam/_models/yoloworld/clip \ --add-data=$(LABELME_PATH)/config/default_config.yaml:labelme/config \ --add-data=$(LABELME_PATH)/icons/*:labelme/icons \ --add-data=$(LABELME_PATH)/translate/*:translate \ --icon=$(LABELME_PATH)/icons/icon-256.png \ --onedir ``` ## Acknowledgement This repo is the fork of [mpitid/pylabelme](https://github.com/mpitid/pylabelme). ================================================ FILE: examples/bbox_detection/README.md ================================================ # Bounding Box Detection Example ## Usage ```bash labelme data_annotated --labels labels.txt ```  ## Convert to VOC-format Dataset ```bash # It generates: # - data_dataset_voc/JPEGImages # - data_dataset_voc/Annotations # - data_dataset_voc/AnnotationsVisualization ./labelme2voc.py data_annotated data_dataset_voc --labels labels.txt ```
Fig1. JPEG image (left), Bounding box annotation visualization (right).
================================================
FILE: examples/bbox_detection/data_annotated/2011_000003.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
191.0,
107.36900369003689
],
[
313.0,
329.36900369003695
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "person",
"points": [
[
365.0,
83.0
],
[
500.0,
333.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
}
],
"imagePath": "2011_000003.jpg",
"imageData": null,
"imageHeight": 338,
"imageWidth": 500
}
================================================
FILE: examples/bbox_detection/data_annotated/2011_000006.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
91.0,
107.0
],
[
240.0,
330.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "person",
"points": [
[
178.0,
110.0
],
[
298.0,
282.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "person",
"points": [
[
254.38461538461536,
115.38461538461539
],
[
369.38461538461536,
292.38461538461536
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "person",
"points": [
[
395.0,
81.0
],
[
447.0,
117.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
}
],
"imagePath": "2011_000006.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/bbox_detection/data_annotated/2011_000025.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "bus",
"points": [
[
84.0,
20.384615384615387
],
[
435.0,
373.38461538461536
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "bus",
"points": [
[
1.0,
99.0
],
[
107.0,
282.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
},
{
"label": "car",
"points": [
[
409.0,
167.0
],
[
500.0,
266.0
]
],
"group_id": null,
"shape_type": "rectangle",
"flags": {}
}
],
"imagePath": "2011_000025.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/bbox_detection/data_dataset_voc/Annotations/2011_000003.xml
================================================
================================================
FILE: examples/classification/data_annotated/0001.json
================================================
{
"version": "4.0.0",
"flags": {
"__ignore__": false,
"cat": true,
"dog": false
},
"shapes": [],
"imagePath": "0001.jpg",
"imageData": null,
"imageHeight": 480,
"imageWidth": 640
}
================================================
FILE: examples/classification/data_annotated/0002.json
================================================
{
"version": "4.0.0",
"flags": {
"__ignore__": false,
"cat": false,
"dog": true
},
"shapes": [],
"imagePath": "0002.jpg",
"imageData": null,
"imageHeight": 480,
"imageWidth": 640
}
================================================
FILE: examples/classification/flags.txt
================================================
__ignore__
cat
dog
================================================
FILE: examples/instance_segmentation/README.md
================================================
# Instance Segmentation Example
## Annotation
```bash
labelme data_annotated --labels labels.txt --validatelabel exact --config '{shift_auto_shape_color: -2}'
labelme data_annotated --labels labels.txt --labelflags '{.*: [occluded, truncated], person: [male]}'
```
## Convert to VOC-format Dataset
```bash
# It generates:
# - data_dataset_voc/JPEGImages
# - data_dataset_voc/SegmentationClass
# - data_dataset_voc/SegmentationClassNpy
# - data_dataset_voc/SegmentationClassVisualization
# - data_dataset_voc/SegmentationObject
# - data_dataset_voc/SegmentationObjectNpy
# - data_dataset_voc/SegmentationObjectVisualization
./labelme2voc.py data_annotated data_dataset_voc --labels labels.txt
```
Fig 1. JPEG image (left), JPEG class label visualization (center), JPEG instance label visualization (right)
Note that the label file contains only very low label values (ex. `0, 4, 14`), and
`255` indicates the `__ignore__` label value (`-1` in the npy file).
You can see the label PNG file by following.
```bash
../tutorial/draw_label_png.py data_dataset_voc/SegmentationClass/2011_000003.png # left
../tutorial/draw_label_png.py data_dataset_voc/SegmentationObject/2011_000003.png # right
```
## Convert to COCO-format Dataset
```bash
# It generates:
# - data_dataset_coco/JPEGImages
# - data_dataset_coco/annotations.json
./labelme2coco.py data_annotated data_dataset_coco --labels labels.txt
```
================================================
FILE: examples/instance_segmentation/data_annotated/2011_000003.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
250.8142292490119,
107.33596837944665
],
[
229.8142292490119,
119.33596837944665
],
[
221.8142292490119,
135.33596837944665
],
[
223.8142292490119,
148.33596837944665
],
[
217.8142292490119,
161.33596837944665
],
[
202.8142292490119,
168.33596837944665
],
[
192.8142292490119,
200.33596837944665
],
[
194.8142292490119,
222.33596837944665
],
[
199.8142292490119,
227.33596837944665
],
[
191.8142292490119,
234.33596837944665
],
[
197.8142292490119,
264.3359683794467
],
[
213.8142292490119,
295.3359683794467
],
[
214.8142292490119,
320.3359683794467
],
[
221.8142292490119,
327.3359683794467
],
[
235.8142292490119,
326.3359683794467
],
[
240.8142292490119,
323.3359683794467
],
[
235.8142292490119,
298.3359683794467
],
[
238.8142292490119,
287.3359683794467
],
[
234.8142292490119,
268.3359683794467
],
[
257.81422924901193,
258.3359683794467
],
[
264.81422924901193,
264.3359683794467
],
[
256.81422924901193,
273.3359683794467
],
[
259.81422924901193,
282.3359683794467
],
[
284.81422924901193,
288.3359683794467
],
[
297.81422924901193,
278.3359683794467
],
[
288.81422924901193,
270.3359683794467
],
[
281.81422924901193,
270.3359683794467
],
[
283.81422924901193,
264.3359683794467
],
[
292.81422924901193,
261.3359683794467
],
[
308.81422924901193,
236.33596837944665
],
[
313.81422924901193,
217.33596837944665
],
[
309.81422924901193,
208.33596837944665
],
[
312.81422924901193,
202.33596837944665
],
[
308.81422924901193,
185.33596837944665
],
[
291.81422924901193,
173.33596837944665
],
[
269.81422924901193,
159.33596837944665
],
[
261.81422924901193,
154.33596837944665
],
[
264.81422924901193,
142.33596837944665
],
[
273.81422924901193,
137.33596837944665
],
[
278.81422924901193,
130.33596837944665
],
[
270.81422924901193,
121.33596837944665
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
482.81422924901193,
87.18098682963114
],
[
468.81422924901193,
92.18098682963114
],
[
460.81422924901193,
112.18098682963114
],
[
460.81422924901193,
129.18098682963114
],
[
444.81422924901193,
139.18098682963114
],
[
419.81422924901193,
155.18098682963114
],
[
410.81422924901193,
165.18098682963114
],
[
403.81422924901193,
170.18098682963114
],
[
394.81422924901193,
172.18098682963114
],
[
386.81422924901193,
170.18098682963114
],
[
386.81422924901193,
186.18098682963114
],
[
392.81422924901193,
184.18098682963114
],
[
410.81422924901193,
189.18098682963114
],
[
414.81422924901193,
194.18098682963114
],
[
437.81422924901193,
191.18098682963114
],
[
434.81422924901193,
206.18098682963114
],
[
390.81422924901193,
197.18098682963114
],
[
386.81422924901193,
197.18098682963114
],
[
387.81422924901193,
210.18098682963114
],
[
381.81422924901193,
214.18098682963114
],
[
372.81422924901193,
214.18098682963114
],
[
372.81422924901193,
218.18098682963114
],
[
400.81422924901193,
272.18098682963114
],
[
389.81422924901193,
274.18098682963114
],
[
389.81422924901193,
276.18098682963114
],
[
403.81422924901193,
284.18098682963114
],
[
444.81422924901193,
285.18098682963114
],
[
443.81422924901193,
261.18098682963114
],
[
426.81422924901193,
246.18098682963114
],
[
462.81422924901193,
258.18098682963114
],
[
474.81422924901193,
272.18098682963114
],
[
477.81422924901193,
282.18098682963114
],
[
473.81422924901193,
291.18098682963114
],
[
471.81422924901193,
298.18098682963114
],
[
472.81422924901193,
319.18098682963114
],
[
480.81422924901193,
334.18098682963114
],
[
494.81422924901193,
337.18098682963114
],
[
498.81422924901193,
331.18098682963114
],
[
494.81422924901193,
310.18098682963114
],
[
499.81422924901193,
299.18098682963114
],
[
499.81422924901193,
92.18098682963114
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
370.81422924901193,
170.33596837944665
],
[
366.81422924901193,
173.33596837944665
],
[
365.81422924901193,
182.33596837944665
],
[
368.81422924901193,
185.33596837944665
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
},
{
"label": "bottle",
"points": [
[
374.81422924901193,
159.33596837944665
],
[
369.81422924901193,
170.33596837944665
],
[
369.81422924901193,
210.33596837944665
],
[
375.81422924901193,
212.33596837944665
],
[
387.81422924901193,
209.33596837944665
],
[
385.81422924901193,
185.33596837944665
],
[
385.81422924901193,
168.33596837944665
],
[
385.81422924901193,
165.33596837944665
],
[
382.81422924901193,
159.33596837944665
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "__ignore__",
"points": [
[
338.81422924901193,
266.3359683794467
],
[
313.81422924901193,
269.3359683794467
],
[
297.81422924901193,
277.3359683794467
],
[
282.81422924901193,
288.3359683794467
],
[
273.81422924901193,
302.3359683794467
],
[
272.81422924901193,
320.3359683794467
],
[
279.81422924901193,
337.3359683794467
],
[
428.81422924901193,
337.3359683794467
],
[
432.81422924901193,
316.3359683794467
],
[
423.81422924901193,
296.3359683794467
],
[
403.81422924901193,
283.3359683794467
],
[
370.81422924901193,
270.3359683794467
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000003.jpg",
"imageData": null,
"imageHeight": 338,
"imageWidth": 500
}
================================================
FILE: examples/instance_segmentation/data_annotated/2011_000006.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
204.936170212766,
108.56382978723406
],
[
183.936170212766,
141.56382978723406
],
[
166.936170212766,
150.56382978723406
],
[
108.93617021276599,
203.56382978723406
],
[
92.93617021276599,
228.56382978723406
],
[
95.93617021276599,
244.56382978723406
],
[
105.93617021276599,
244.56382978723406
],
[
116.93617021276599,
223.56382978723406
],
[
163.936170212766,
187.56382978723406
],
[
147.936170212766,
212.56382978723406
],
[
117.93617021276599,
222.56382978723406
],
[
108.93617021276599,
243.56382978723406
],
[
100.93617021276599,
325.56382978723406
],
[
135.936170212766,
329.56382978723406
],
[
148.936170212766,
319.56382978723406
],
[
150.936170212766,
295.56382978723406
],
[
169.936170212766,
272.56382978723406
],
[
171.936170212766,
249.56382978723406
],
[
178.936170212766,
246.56382978723406
],
[
186.936170212766,
225.56382978723406
],
[
214.936170212766,
219.56382978723406
],
[
242.936170212766,
157.56382978723406
],
[
228.936170212766,
146.56382978723406
],
[
228.936170212766,
125.56382978723406
],
[
216.936170212766,
112.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
271.936170212766,
109.56382978723406
],
[
249.936170212766,
110.56382978723406
],
[
244.936170212766,
150.56382978723406
],
[
215.936170212766,
219.56382978723406
],
[
208.936170212766,
245.56382978723406
],
[
214.936170212766,
220.56382978723406
],
[
188.936170212766,
227.56382978723406
],
[
170.936170212766,
246.56382978723406
],
[
170.936170212766,
275.56382978723406
],
[
221.936170212766,
278.56382978723406
],
[
233.936170212766,
259.56382978723406
],
[
246.936170212766,
253.56382978723406
],
[
245.936170212766,
256.56382978723406
],
[
242.936170212766,
251.56382978723406
],
[
262.936170212766,
256.56382978723406
],
[
304.936170212766,
226.56382978723406
],
[
297.936170212766,
199.56382978723406
],
[
308.936170212766,
164.56382978723406
],
[
296.936170212766,
148.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
308.936170212766,
115.56382978723406
],
[
298.936170212766,
145.56382978723406
],
[
309.936170212766,
166.56382978723406
],
[
297.936170212766,
200.56382978723406
],
[
305.936170212766,
228.56382978723406
],
[
262.936170212766,
258.56382978723406
],
[
252.936170212766,
284.56382978723406
],
[
272.936170212766,
291.56382978723406
],
[
281.936170212766,
250.56382978723406
],
[
326.936170212766,
235.56382978723406
],
[
351.936170212766,
239.56382978723406
],
[
365.936170212766,
223.56382978723406
],
[
371.936170212766,
187.56382978723406
],
[
353.936170212766,
168.56382978723406
],
[
344.936170212766,
143.56382978723406
],
[
336.936170212766,
115.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "chair",
"points": [
[
309.7054009819968,
242.94844517184941
],
[
282.7054009819968,
251.94844517184941
],
[
271.7054009819968,
287.9484451718494
],
[
175.70540098199677,
275.9484451718494
],
[
149.70540098199677,
296.9484451718494
],
[
151.70540098199677,
319.9484451718494
],
[
160.70540098199677,
328.9484451718494
],
[
165.54250204582655,
375.38461538461536
],
[
486.7054009819968,
373.9484451718494
],
[
498.7054009819968,
336.9484451718494
],
[
498.7054009819968,
202.94844517184941
],
[
454.7054009819968,
193.94844517184941
],
[
435.7054009819968,
212.94844517184941
],
[
368.7054009819968,
224.94844517184941
],
[
351.7054009819968,
241.94844517184941
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
425.936170212766,
82.56382978723406
],
[
404.936170212766,
109.56382978723406
],
[
400.936170212766,
114.56382978723406
],
[
437.936170212766,
114.56382978723406
],
[
448.936170212766,
102.56382978723406
],
[
446.936170212766,
91.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "__ignore__",
"points": [
[
457.936170212766,
85.56382978723406
],
[
439.936170212766,
117.56382978723406
],
[
477.936170212766,
117.56382978723406
],
[
474.936170212766,
87.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
183.936170212766,
140.56382978723406
],
[
125.93617021276599,
140.56382978723406
],
[
110.93617021276599,
187.56382978723406
],
[
22.936170212765987,
199.56382978723406
],
[
18.936170212765987,
218.56382978723406
],
[
22.936170212765987,
234.56382978723406
],
[
93.93617021276599,
239.56382978723406
],
[
91.93617021276599,
229.56382978723406
],
[
110.93617021276599,
203.56382978723406
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
103.93617021276599,
290.56382978723406
],
[
58.93617021276599,
303.56382978723406
],
[
97.93617021276599,
311.56382978723406
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
348.936170212766,
146.56382978723406
],
[
472.936170212766,
149.56382978723406
],
[
477.936170212766,
162.56382978723406
],
[
471.936170212766,
196.56382978723406
],
[
453.936170212766,
192.56382978723406
],
[
434.936170212766,
213.56382978723406
],
[
368.936170212766,
226.56382978723406
],
[
375.936170212766,
187.56382978723406
],
[
353.936170212766,
164.56382978723406
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
246.936170212766,
252.56382978723406
],
[
219.936170212766,
277.56382978723406
],
[
254.936170212766,
287.56382978723406
],
[
261.936170212766,
256.56382978723406
]
],
"group_id": 0,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000006.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/instance_segmentation/data_annotated/2011_000025.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "bus",
"points": [
[
260.936170212766,
23.33306055646483
],
[
193.936170212766,
20.33306055646483
],
[
124.93617021276599,
40.33306055646483
],
[
89.93617021276599,
102.33306055646483
],
[
81.93617021276599,
151.33306055646483
],
[
108.93617021276599,
146.33306055646483
],
[
88.93617021276599,
245.33306055646483
],
[
89.93617021276599,
323.33306055646483
],
[
116.93617021276599,
368.33306055646483
],
[
158.936170212766,
369.33306055646483
],
[
165.936170212766,
338.33306055646483
],
[
347.936170212766,
336.33306055646483
],
[
349.936170212766,
370.33306055646483
],
[
391.936170212766,
374.33306055646483
],
[
403.936170212766,
336.33306055646483
],
[
425.936170212766,
333.33306055646483
],
[
421.936170212766,
282.33306055646483
],
[
428.936170212766,
253.33306055646483
],
[
428.936170212766,
237.33306055646483
],
[
409.936170212766,
221.33306055646483
],
[
409.936170212766,
151.33306055646483
],
[
430.936170212766,
144.33306055646483
],
[
433.936170212766,
113.33306055646483
],
[
431.936170212766,
97.33306055646483
],
[
408.936170212766,
91.33306055646483
],
[
395.936170212766,
51.33306055646483
],
[
338.936170212766,
26.33306055646483
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "bus",
"points": [
[
88.93617021276599,
115.56382978723406
],
[
0.9361702127659877,
96.56382978723406
],
[
0.0,
251.968085106388
],
[
0.9361702127659877,
265.56382978723406
],
[
27.936170212765987,
265.56382978723406
],
[
29.936170212765987,
283.56382978723406
],
[
63.93617021276599,
281.56382978723406
],
[
89.93617021276599,
252.56382978723406
],
[
100.93617021276599,
183.56382978723406
],
[
108.93617021276599,
145.56382978723406
],
[
81.93617021276599,
151.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
413.936170212766,
168.94844517184944
],
[
497.936170212766,
168.94844517184944
],
[
497.936170212766,
256.94844517184936
],
[
431.936170212766,
258.94844517184936
],
[
430.936170212766,
236.94844517184944
],
[
408.936170212766,
218.94844517184944
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000025.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/instance_segmentation/data_dataset_coco/annotations.json
================================================
{"info": {"description": null, "url": null, "version": null, "year": 2020, "contributor": null, "date_created": "2020-01-26 05:46:30.244442"}, "licenses": [{"url": null, "id": 0, "name": null}], "images": [{"license": 0, "url": null, "file_name": "JPEGImages/2011_000003.jpg", "height": 338, "width": 500, "date_captured": null, "id": 0}, {"license": 0, "url": null, "file_name": "JPEGImages/2011_000025.jpg", "height": 375, "width": 500, "date_captured": null, "id": 1}, {"license": 0, "url": null, "file_name": "JPEGImages/2011_000006.jpg", "height": 375, "width": 500, "date_captured": null, "id": 2}], "type": "instances", "annotations": [{"id": 0, "image_id": 0, "category_id": 15, "segmentation": [[250.8142292490119, 107.33596837944665, 229.8142292490119, 119.33596837944665, 221.8142292490119, 135.33596837944665, 223.8142292490119, 148.33596837944665, 217.8142292490119, 161.33596837944665, 202.8142292490119, 168.33596837944665, 192.8142292490119, 200.33596837944665, 194.8142292490119, 222.33596837944665, 199.8142292490119, 227.33596837944665, 191.8142292490119, 234.33596837944665, 197.8142292490119, 264.3359683794467, 213.8142292490119, 295.3359683794467, 214.8142292490119, 320.3359683794467, 221.8142292490119, 327.3359683794467, 235.8142292490119, 326.3359683794467, 240.8142292490119, 323.3359683794467, 235.8142292490119, 298.3359683794467, 238.8142292490119, 287.3359683794467, 234.8142292490119, 268.3359683794467, 257.81422924901193, 258.3359683794467, 264.81422924901193, 264.3359683794467, 256.81422924901193, 273.3359683794467, 259.81422924901193, 282.3359683794467, 284.81422924901193, 288.3359683794467, 297.81422924901193, 278.3359683794467, 288.81422924901193, 270.3359683794467, 281.81422924901193, 270.3359683794467, 283.81422924901193, 264.3359683794467, 292.81422924901193, 261.3359683794467, 308.81422924901193, 236.33596837944665, 313.81422924901193, 217.33596837944665, 309.81422924901193, 208.33596837944665, 312.81422924901193, 202.33596837944665, 308.81422924901193, 185.33596837944665, 291.81422924901193, 173.33596837944665, 269.81422924901193, 159.33596837944665, 261.81422924901193, 154.33596837944665, 264.81422924901193, 142.33596837944665, 273.81422924901193, 137.33596837944665, 278.81422924901193, 130.33596837944665, 270.81422924901193, 121.33596837944665]], "area": 15689.0, "bbox": [191.0, 107.0, 123.0, 221.0], "iscrowd": 0}, {"id": 1, "image_id": 0, "category_id": 15, "segmentation": [[482.81422924901193, 87.18098682963114, 468.81422924901193, 92.18098682963114, 460.81422924901193, 112.18098682963114, 460.81422924901193, 129.18098682963114, 444.81422924901193, 139.18098682963114, 419.81422924901193, 155.18098682963114, 410.81422924901193, 165.18098682963114, 403.81422924901193, 170.18098682963114, 394.81422924901193, 172.18098682963114, 386.81422924901193, 170.18098682963114, 386.81422924901193, 186.18098682963114, 392.81422924901193, 184.18098682963114, 410.81422924901193, 189.18098682963114, 414.81422924901193, 194.18098682963114, 437.81422924901193, 191.18098682963114, 434.81422924901193, 206.18098682963114, 390.81422924901193, 197.18098682963114, 386.81422924901193, 197.18098682963114, 387.81422924901193, 210.18098682963114, 381.81422924901193, 214.18098682963114, 372.81422924901193, 214.18098682963114, 372.81422924901193, 218.18098682963114, 400.81422924901193, 272.18098682963114, 389.81422924901193, 274.18098682963114, 389.81422924901193, 276.18098682963114, 403.81422924901193, 284.18098682963114, 444.81422924901193, 285.18098682963114, 443.81422924901193, 261.18098682963114, 426.81422924901193, 246.18098682963114, 462.81422924901193, 258.18098682963114, 474.81422924901193, 272.18098682963114, 477.81422924901193, 282.18098682963114, 473.81422924901193, 291.18098682963114, 471.81422924901193, 298.18098682963114, 472.81422924901193, 319.18098682963114, 480.81422924901193, 334.18098682963114, 494.81422924901193, 337.18098682963114, 498.81422924901193, 331.18098682963114, 494.81422924901193, 310.18098682963114, 499.81422924901193, 299.18098682963114, 499.81422924901193, 92.18098682963114], [370.81422924901193, 170.33596837944665, 366.81422924901193, 173.33596837944665, 365.81422924901193, 182.33596837944665, 368.81422924901193, 185.33596837944665]], "area": 17254.0, "bbox": [365.0, 87.0, 135.0, 251.0], "iscrowd": 0}, {"id": 2, "image_id": 0, "category_id": 5, "segmentation": [[374.81422924901193, 159.33596837944665, 369.81422924901193, 170.33596837944665, 369.81422924901193, 210.33596837944665, 375.81422924901193, 212.33596837944665, 387.81422924901193, 209.33596837944665, 385.81422924901193, 185.33596837944665, 385.81422924901193, 168.33596837944665, 385.81422924901193, 165.33596837944665, 382.81422924901193, 159.33596837944665]], "area": 873.0, "bbox": [369.0, 159.0, 19.0, 54.0], "iscrowd": 0}, {"id": 3, "image_id": 1, "category_id": 6, "segmentation": [[260.936170212766, 23.33306055646483, 193.936170212766, 20.33306055646483, 124.93617021276599, 40.33306055646483, 89.93617021276599, 102.33306055646483, 81.93617021276599, 151.33306055646483, 108.93617021276599, 146.33306055646483, 88.93617021276599, 245.33306055646483, 89.93617021276599, 323.33306055646483, 116.93617021276599, 368.33306055646483, 158.936170212766, 369.33306055646483, 165.936170212766, 338.33306055646483, 347.936170212766, 336.33306055646483, 349.936170212766, 370.33306055646483, 391.936170212766, 374.33306055646483, 403.936170212766, 336.33306055646483, 425.936170212766, 333.33306055646483, 421.936170212766, 282.33306055646483, 428.936170212766, 253.33306055646483, 428.936170212766, 237.33306055646483, 409.936170212766, 221.33306055646483, 409.936170212766, 151.33306055646483, 430.936170212766, 144.33306055646483, 433.936170212766, 113.33306055646483, 431.936170212766, 97.33306055646483, 408.936170212766, 91.33306055646483, 395.936170212766, 51.33306055646483, 338.936170212766, 26.33306055646483]], "area": 102701.0, "bbox": [81.0, 20.0, 353.0, 355.0], "iscrowd": 0}, {"id": 4, "image_id": 1, "category_id": 6, "segmentation": [[88.93617021276599, 115.56382978723406, 0.9361702127659877, 96.56382978723406, 0.0, 251.968085106388, 0.9361702127659877, 265.56382978723406, 27.936170212765987, 265.56382978723406, 29.936170212765987, 283.56382978723406, 63.93617021276599, 281.56382978723406, 89.93617021276599, 252.56382978723406, 100.93617021276599, 183.56382978723406, 108.93617021276599, 145.56382978723406, 81.93617021276599, 151.56382978723406]], "area": 15781.0, "bbox": [0.0, 96.0, 109.0, 188.0], "iscrowd": 0}, {"id": 5, "image_id": 1, "category_id": 7, "segmentation": [[413.936170212766, 168.94844517184944, 497.936170212766, 168.94844517184944, 497.936170212766, 256.94844517184936, 431.936170212766, 258.94844517184936, 430.936170212766, 236.94844517184944, 408.936170212766, 218.94844517184944]], "area": 7256.0, "bbox": [408.0, 168.0, 90.0, 91.0], "iscrowd": 0}, {"id": 6, "image_id": 2, "category_id": 15, "segmentation": [[204.936170212766, 108.56382978723406, 183.936170212766, 141.56382978723406, 166.936170212766, 150.56382978723406, 108.93617021276599, 203.56382978723406, 92.93617021276599, 228.56382978723406, 95.93617021276599, 244.56382978723406, 105.93617021276599, 244.56382978723406, 116.93617021276599, 223.56382978723406, 163.936170212766, 187.56382978723406, 147.936170212766, 212.56382978723406, 117.93617021276599, 222.56382978723406, 108.93617021276599, 243.56382978723406, 100.93617021276599, 325.56382978723406, 135.936170212766, 329.56382978723406, 148.936170212766, 319.56382978723406, 150.936170212766, 295.56382978723406, 169.936170212766, 272.56382978723406, 171.936170212766, 249.56382978723406, 178.936170212766, 246.56382978723406, 186.936170212766, 225.56382978723406, 214.936170212766, 219.56382978723406, 242.936170212766, 157.56382978723406, 228.936170212766, 146.56382978723406, 228.936170212766, 125.56382978723406, 216.936170212766, 112.56382978723406]], "area": 15203.0, "bbox": [92.0, 108.0, 151.0, 222.0], "iscrowd": 0}, {"id": 7, "image_id": 2, "category_id": 15, "segmentation": [[271.936170212766, 109.56382978723406, 249.936170212766, 110.56382978723406, 244.936170212766, 150.56382978723406, 215.936170212766, 219.56382978723406, 208.936170212766, 245.56382978723406, 214.936170212766, 220.56382978723406, 188.936170212766, 227.56382978723406, 170.936170212766, 246.56382978723406, 170.936170212766, 275.56382978723406, 221.936170212766, 278.56382978723406, 233.936170212766, 259.56382978723406, 246.936170212766, 253.56382978723406, 245.936170212766, 256.56382978723406, 242.936170212766, 251.56382978723406, 262.936170212766, 256.56382978723406, 304.936170212766, 226.56382978723406, 297.936170212766, 199.56382978723406, 308.936170212766, 164.56382978723406, 296.936170212766, 148.56382978723406]], "area": 11735.0, "bbox": [170.0, 109.0, 139.0, 170.0], "iscrowd": 0}, {"id": 8, "image_id": 2, "category_id": 15, "segmentation": [[308.936170212766, 115.56382978723406, 298.936170212766, 145.56382978723406, 309.936170212766, 166.56382978723406, 297.936170212766, 200.56382978723406, 305.936170212766, 228.56382978723406, 262.936170212766, 258.56382978723406, 252.936170212766, 284.56382978723406, 272.936170212766, 291.56382978723406, 281.936170212766, 250.56382978723406, 326.936170212766, 235.56382978723406, 351.936170212766, 239.56382978723406, 365.936170212766, 223.56382978723406, 371.936170212766, 187.56382978723406, 353.936170212766, 168.56382978723406, 344.936170212766, 143.56382978723406, 336.936170212766, 115.56382978723406]], "area": 7597.0, "bbox": [252.0, 115.0, 120.0, 177.0], "iscrowd": 0}, {"id": 9, "image_id": 2, "category_id": 9, "segmentation": [[309.7054009819968, 242.94844517184941, 282.7054009819968, 251.94844517184941, 271.7054009819968, 287.9484451718494, 175.70540098199677, 275.9484451718494, 149.70540098199677, 296.9484451718494, 151.70540098199677, 319.9484451718494, 160.70540098199677, 328.9484451718494, 165.54250204582655, 375.38461538461536, 486.7054009819968, 373.9484451718494, 498.7054009819968, 336.9484451718494, 498.7054009819968, 202.94844517184941, 454.7054009819968, 193.94844517184941, 435.7054009819968, 212.94844517184941, 368.7054009819968, 224.94844517184941, 351.7054009819968, 241.94844517184941]], "area": 44532.0, "bbox": [149.0, 193.0, 350.0, 182.0], "iscrowd": 0}, {"id": 10, "image_id": 2, "category_id": 15, "segmentation": [[425.936170212766, 82.56382978723406, 404.936170212766, 109.56382978723406, 400.936170212766, 114.56382978723406, 437.936170212766, 114.56382978723406, 448.936170212766, 102.56382978723406, 446.936170212766, 91.56382978723406]], "area": 996.0, "bbox": [400.0, 82.0, 49.0, 33.0], "iscrowd": 0}, {"id": 11, "image_id": 2, "category_id": 18, "segmentation": [[183.936170212766, 140.56382978723406, 125.93617021276599, 140.56382978723406, 110.93617021276599, 187.56382978723406, 22.936170212765987, 199.56382978723406, 18.936170212765987, 218.56382978723406, 22.936170212765987, 234.56382978723406, 93.93617021276599, 239.56382978723406, 91.93617021276599, 229.56382978723406, 110.93617021276599, 203.56382978723406], [103.93617021276599, 290.56382978723406, 58.93617021276599, 303.56382978723406, 97.93617021276599, 311.56382978723406], [348.936170212766, 146.56382978723406, 472.936170212766, 149.56382978723406, 477.936170212766, 162.56382978723406, 471.936170212766, 196.56382978723406, 453.936170212766, 192.56382978723406, 434.936170212766, 213.56382978723406, 368.936170212766, 226.56382978723406, 375.936170212766, 187.56382978723406, 353.936170212766, 164.56382978723406], [246.936170212766, 252.56382978723406, 219.936170212766, 277.56382978723406, 254.936170212766, 287.56382978723406, 261.936170212766, 256.56382978723406]], "area": 14001.0, "bbox": [18.0, 140.0, 460.0, 172.0], "iscrowd": 0}], "categories": [{"supercategory": null, "id": 0, "name": "_background_"}, {"supercategory": null, "id": 1, "name": "aeroplane"}, {"supercategory": null, "id": 2, "name": "bicycle"}, {"supercategory": null, "id": 3, "name": "bird"}, {"supercategory": null, "id": 4, "name": "boat"}, {"supercategory": null, "id": 5, "name": "bottle"}, {"supercategory": null, "id": 6, "name": "bus"}, {"supercategory": null, "id": 7, "name": "car"}, {"supercategory": null, "id": 8, "name": "cat"}, {"supercategory": null, "id": 9, "name": "chair"}, {"supercategory": null, "id": 10, "name": "cow"}, {"supercategory": null, "id": 11, "name": "diningtable"}, {"supercategory": null, "id": 12, "name": "dog"}, {"supercategory": null, "id": 13, "name": "horse"}, {"supercategory": null, "id": 14, "name": "motorbike"}, {"supercategory": null, "id": 15, "name": "person"}, {"supercategory": null, "id": 16, "name": "potted plant"}, {"supercategory": null, "id": 17, "name": "sheep"}, {"supercategory": null, "id": 18, "name": "sofa"}, {"supercategory": null, "id": 19, "name": "train"}, {"supercategory": null, "id": 20, "name": "tv/monitor"}]}
================================================
FILE: examples/instance_segmentation/data_dataset_voc/class_names.txt
================================================
_background_
aeroplane
bicycle
bird
boat
bottle
bus
car
cat
chair
cow
diningtable
dog
horse
motorbike
person
potted plant
sheep
sofa
train
tv/monitor
================================================
FILE: examples/instance_segmentation/labelme2coco.py
================================================
#!/usr/bin/env python
import argparse
import collections
import datetime
import glob
import json
import os
import os.path as osp
import sys
import uuid
import imgviz
import numpy as np
import labelme
try:
import pycocotools.mask # type: ignore
except ImportError:
print("Please install pycocotools:\n\n pip install pycocotools\n")
sys.exit(1)
def main():
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
parser.add_argument("input_dir", help="input annotated directory")
parser.add_argument("output_dir", help="output dataset directory")
parser.add_argument("--labels", help="labels file", required=True)
parser.add_argument("--noviz", help="no visualization", action="store_true")
args = parser.parse_args()
if osp.exists(args.output_dir):
print("Output directory already exists:", args.output_dir)
sys.exit(1)
os.makedirs(args.output_dir)
os.makedirs(osp.join(args.output_dir, "JPEGImages"))
if not args.noviz:
os.makedirs(osp.join(args.output_dir, "Visualization"))
print("Creating dataset:", args.output_dir)
now = datetime.datetime.now()
data = dict(
info=dict(
description=None,
url=None,
version=None,
year=now.year,
contributor=None,
date_created=now.strftime("%Y-%m-%d %H:%M:%S.%f"),
),
licenses=[
dict(
url=None,
id=0,
name=None,
)
],
type="instances",
)
data["images"] = [] # license, url, file_name, height, width, date_captured, id
data["categories"] = [] # supercategory, id, name
data[
"annotations"
] = [] # segmentation, area, iscrowd, image_id, bbox, category_id, id
class_name_to_id = {}
for i, line in enumerate(open(args.labels).readlines()):
class_id = i - 1 # starts with -1
class_name = line.strip()
if class_id == -1:
assert class_name == "__ignore__"
continue
class_name_to_id[class_name] = class_id
data["categories"].append(
dict(
supercategory=None,
id=class_id,
name=class_name,
)
)
out_ann_file = osp.join(args.output_dir, "annotations.json")
label_files = glob.glob(osp.join(args.input_dir, "*.json"))
for image_id, filename in enumerate(label_files):
print("Generating dataset from:", filename)
label_file = labelme.LabelFile(filename=filename)
base = osp.splitext(osp.basename(filename))[0]
out_img_file = osp.join(args.output_dir, "JPEGImages", f"{base}.jpg")
img = labelme.utils.img_data_to_arr(label_file.imageData)
if img.ndim == 3 and img.shape[2] == 4:
img = imgviz.rgba2rgb(img)
imgviz.io.imsave(out_img_file, img)
data["images"].append(
dict(
license=0,
url=None,
file_name=osp.relpath(out_img_file, osp.dirname(out_ann_file)),
height=img.shape[0],
width=img.shape[1],
date_captured=None,
id=image_id,
)
)
masks = {} # for area
segmentations = collections.defaultdict(list) # for segmentation
for shape in label_file.shapes:
points: list[list[int | float]] = shape["points"]
label = shape["label"]
group_id = shape.get("group_id")
shape_type = shape.get("shape_type", "polygon")
mask = labelme.utils.shape_to_mask(img.shape[:2], points, shape_type)
if group_id is None:
group_id = uuid.uuid1()
instance = (label, group_id)
if instance in masks:
masks[instance] = masks[instance] | mask
else:
masks[instance] = mask
points_coco: list[int | float]
if shape_type == "rectangle":
(x1, y1), (x2, y2) = points
x1, x2 = sorted([x1, x2])
y1, y2 = sorted([y1, y2])
points_coco = [x1, y1, x2, y1, x2, y2, x1, y2]
if shape_type == "circle":
(x1, y1), (x2, y2) = points
r = np.linalg.norm([x2 - x1, y2 - y1])
# r(1-cos(a/2)) 
Fig 1. JPEG image (left), PNG label (center), JPEG label visualization (right)
Note that the label file contains only very low label values (ex. `0, 4, 14`), and
`255` indicates the `__ignore__` label value (`-1` in the npy file).
You can see the label PNG file by following.
```bash
../tutorial/draw_label_png.py data_dataset_voc/SegmentationClass/2011_000003.png
```
================================================
FILE: examples/semantic_segmentation/data_annotated/2011_000003.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
250.8142292490119,
106.96696468940974
],
[
229.8142292490119,
118.96696468940974
],
[
221.8142292490119,
134.96696468940974
],
[
223.8142292490119,
147.96696468940974
],
[
217.8142292490119,
160.96696468940974
],
[
202.8142292490119,
167.96696468940974
],
[
192.8142292490119,
199.96696468940974
],
[
194.8142292490119,
221.96696468940974
],
[
199.8142292490119,
226.96696468940974
],
[
191.8142292490119,
233.96696468940974
],
[
197.8142292490119,
263.9669646894098
],
[
213.8142292490119,
294.9669646894098
],
[
214.8142292490119,
319.9669646894098
],
[
221.8142292490119,
326.9669646894098
],
[
235.8142292490119,
325.9669646894098
],
[
240.8142292490119,
322.9669646894098
],
[
235.8142292490119,
297.9669646894098
],
[
238.8142292490119,
286.9669646894098
],
[
234.8142292490119,
267.9669646894098
],
[
257.81422924901193,
257.9669646894098
],
[
264.81422924901193,
263.9669646894098
],
[
256.81422924901193,
272.9669646894098
],
[
259.81422924901193,
281.9669646894098
],
[
284.81422924901193,
287.9669646894098
],
[
297.81422924901193,
277.9669646894098
],
[
288.81422924901193,
269.9669646894098
],
[
281.81422924901193,
269.9669646894098
],
[
283.81422924901193,
263.9669646894098
],
[
292.81422924901193,
260.9669646894098
],
[
308.81422924901193,
235.96696468940974
],
[
313.81422924901193,
216.96696468940974
],
[
309.81422924901193,
207.96696468940974
],
[
312.81422924901193,
201.96696468940974
],
[
308.81422924901193,
184.96696468940974
],
[
291.81422924901193,
172.96696468940974
],
[
269.81422924901193,
158.96696468940974
],
[
261.81422924901193,
153.96696468940974
],
[
264.81422924901193,
141.96696468940974
],
[
273.81422924901193,
136.96696468940974
],
[
278.81422924901193,
129.96696468940974
],
[
270.81422924901193,
120.96696468940974
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
482.81422924901193,
85.33596837944665
],
[
468.81422924901193,
90.33596837944665
],
[
460.81422924901193,
110.33596837944665
],
[
460.81422924901193,
127.33596837944665
],
[
444.81422924901193,
137.33596837944665
],
[
419.81422924901193,
153.33596837944665
],
[
410.81422924901193,
163.33596837944665
],
[
403.81422924901193,
168.33596837944665
],
[
394.81422924901193,
170.33596837944665
],
[
386.81422924901193,
168.33596837944665
],
[
386.81422924901193,
184.33596837944665
],
[
392.81422924901193,
182.33596837944665
],
[
410.81422924901193,
187.33596837944665
],
[
414.81422924901193,
192.33596837944665
],
[
437.81422924901193,
189.33596837944665
],
[
434.81422924901193,
204.33596837944665
],
[
390.81422924901193,
195.33596837944665
],
[
386.81422924901193,
195.33596837944665
],
[
387.81422924901193,
208.33596837944665
],
[
381.81422924901193,
212.33596837944665
],
[
372.81422924901193,
212.33596837944665
],
[
372.81422924901193,
216.33596837944665
],
[
400.81422924901193,
270.3359683794467
],
[
389.81422924901193,
272.3359683794467
],
[
389.81422924901193,
274.3359683794467
],
[
403.81422924901193,
282.3359683794467
],
[
444.81422924901193,
283.3359683794467
],
[
443.81422924901193,
259.3359683794467
],
[
426.81422924901193,
244.33596837944665
],
[
462.81422924901193,
256.3359683794467
],
[
474.81422924901193,
270.3359683794467
],
[
477.81422924901193,
280.3359683794467
],
[
473.81422924901193,
289.3359683794467
],
[
471.81422924901193,
296.3359683794467
],
[
472.81422924901193,
317.3359683794467
],
[
480.81422924901193,
332.3359683794467
],
[
494.81422924901193,
335.3359683794467
],
[
498.81422924901193,
329.3359683794467
],
[
494.81422924901193,
308.3359683794467
],
[
499.81422924901193,
297.3359683794467
],
[
499.81422924901193,
90.33596837944665
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
370.81422924901193,
170.33596837944665
],
[
366.81422924901193,
173.33596837944665
],
[
365.81422924901193,
182.33596837944665
],
[
368.81422924901193,
185.33596837944665
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "bottle",
"points": [
[
374.81422924901193,
159.33596837944665
],
[
369.81422924901193,
170.33596837944665
],
[
369.81422924901193,
210.33596837944665
],
[
375.81422924901193,
212.33596837944665
],
[
387.81422924901193,
209.33596837944665
],
[
385.81422924901193,
185.33596837944665
],
[
385.81422924901193,
168.33596837944665
],
[
385.81422924901193,
165.33596837944665
],
[
382.81422924901193,
159.33596837944665
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "__ignore__",
"points": [
[
338.81422924901193,
266.3359683794467
],
[
313.81422924901193,
269.3359683794467
],
[
297.81422924901193,
277.3359683794467
],
[
282.81422924901193,
288.3359683794467
],
[
273.81422924901193,
302.3359683794467
],
[
272.81422924901193,
320.3359683794467
],
[
279.81422924901193,
337.3359683794467
],
[
428.81422924901193,
337.3359683794467
],
[
432.81422924901193,
316.3359683794467
],
[
423.81422924901193,
296.3359683794467
],
[
403.81422924901193,
283.3359683794467
],
[
370.81422924901193,
270.3359683794467
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000003.jpg",
"imageData": null,
"imageHeight": 338,
"imageWidth": 500
}
================================================
FILE: examples/semantic_segmentation/data_annotated/2011_000006.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "person",
"points": [
[
204.936170212766,
108.56382978723406
],
[
183.936170212766,
141.56382978723406
],
[
166.936170212766,
150.56382978723406
],
[
108.93617021276599,
203.56382978723406
],
[
92.93617021276599,
228.56382978723406
],
[
95.93617021276599,
244.56382978723406
],
[
105.93617021276599,
244.56382978723406
],
[
116.93617021276599,
223.56382978723406
],
[
163.936170212766,
187.56382978723406
],
[
147.936170212766,
212.56382978723406
],
[
117.93617021276599,
222.56382978723406
],
[
108.93617021276599,
243.56382978723406
],
[
100.93617021276599,
325.56382978723406
],
[
135.936170212766,
329.56382978723406
],
[
148.936170212766,
319.56382978723406
],
[
150.936170212766,
295.56382978723406
],
[
169.936170212766,
272.56382978723406
],
[
171.936170212766,
249.56382978723406
],
[
178.936170212766,
246.56382978723406
],
[
186.936170212766,
225.56382978723406
],
[
214.936170212766,
219.56382978723406
],
[
242.936170212766,
157.56382978723406
],
[
228.936170212766,
146.56382978723406
],
[
228.936170212766,
125.56382978723406
],
[
216.936170212766,
112.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
271.936170212766,
109.56382978723406
],
[
249.936170212766,
110.56382978723406
],
[
244.936170212766,
150.56382978723406
],
[
215.936170212766,
219.56382978723406
],
[
208.936170212766,
245.56382978723406
],
[
214.936170212766,
220.56382978723406
],
[
188.936170212766,
227.56382978723406
],
[
170.936170212766,
246.56382978723406
],
[
170.936170212766,
275.56382978723406
],
[
221.936170212766,
278.56382978723406
],
[
233.936170212766,
259.56382978723406
],
[
246.936170212766,
253.56382978723406
],
[
245.936170212766,
256.56382978723406
],
[
242.936170212766,
251.56382978723406
],
[
262.936170212766,
256.56382978723406
],
[
304.936170212766,
226.56382978723406
],
[
297.936170212766,
199.56382978723406
],
[
308.936170212766,
164.56382978723406
],
[
296.936170212766,
148.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
308.936170212766,
115.56382978723406
],
[
298.936170212766,
145.56382978723406
],
[
309.936170212766,
166.56382978723406
],
[
297.936170212766,
200.56382978723406
],
[
305.936170212766,
228.56382978723406
],
[
262.936170212766,
258.56382978723406
],
[
252.936170212766,
284.56382978723406
],
[
272.936170212766,
291.56382978723406
],
[
281.936170212766,
250.56382978723406
],
[
326.936170212766,
235.56382978723406
],
[
351.936170212766,
239.56382978723406
],
[
365.936170212766,
223.56382978723406
],
[
371.936170212766,
187.56382978723406
],
[
353.936170212766,
168.56382978723406
],
[
344.936170212766,
143.56382978723406
],
[
336.936170212766,
115.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "chair",
"points": [
[
308.936170212766,
243.33306055646483
],
[
281.936170212766,
252.33306055646483
],
[
270.936170212766,
288.33306055646483
],
[
174.936170212766,
276.33306055646483
],
[
148.936170212766,
297.33306055646483
],
[
150.936170212766,
320.33306055646483
],
[
159.936170212766,
329.33306055646483
],
[
164.77327127659578,
375.7692307692308
],
[
485.936170212766,
374.33306055646483
],
[
497.936170212766,
337.33306055646483
],
[
497.936170212766,
203.33306055646483
],
[
453.936170212766,
194.33306055646483
],
[
434.936170212766,
213.33306055646483
],
[
367.936170212766,
225.33306055646483
],
[
350.936170212766,
242.33306055646483
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "person",
"points": [
[
425.936170212766,
82.56382978723406
],
[
404.936170212766,
109.56382978723406
],
[
400.936170212766,
114.56382978723406
],
[
437.936170212766,
114.56382978723406
],
[
448.936170212766,
102.56382978723406
],
[
446.936170212766,
91.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "__ignore__",
"points": [
[
457.936170212766,
85.56382978723406
],
[
439.936170212766,
117.56382978723406
],
[
477.936170212766,
117.56382978723406
],
[
474.936170212766,
87.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
183.936170212766,
140.56382978723406
],
[
125.93617021276599,
140.56382978723406
],
[
110.93617021276599,
187.56382978723406
],
[
22.936170212765987,
199.56382978723406
],
[
18.936170212765987,
218.56382978723406
],
[
22.936170212765987,
234.56382978723406
],
[
93.93617021276599,
239.56382978723406
],
[
91.93617021276599,
229.56382978723406
],
[
110.93617021276599,
203.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
103.93617021276599,
290.56382978723406
],
[
58.93617021276599,
303.56382978723406
],
[
97.93617021276599,
311.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
348.936170212766,
146.56382978723406
],
[
472.936170212766,
149.56382978723406
],
[
477.936170212766,
162.56382978723406
],
[
471.936170212766,
196.56382978723406
],
[
453.936170212766,
192.56382978723406
],
[
434.936170212766,
213.56382978723406
],
[
368.936170212766,
226.56382978723406
],
[
375.936170212766,
187.56382978723406
],
[
353.936170212766,
164.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "sofa",
"points": [
[
246.936170212766,
252.56382978723406
],
[
219.936170212766,
277.56382978723406
],
[
254.936170212766,
287.56382978723406
],
[
261.936170212766,
256.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000006.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/semantic_segmentation/data_annotated/2011_000025.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "bus",
"points": [
[
260.936170212766,
22.948445171849443
],
[
193.936170212766,
19.948445171849443
],
[
124.93617021276599,
39.94844517184944
],
[
89.93617021276599,
101.94844517184944
],
[
81.93617021276599,
150.94844517184944
],
[
108.93617021276599,
145.94844517184944
],
[
88.93617021276599,
244.94844517184944
],
[
89.93617021276599,
322.94844517184936
],
[
116.93617021276599,
367.94844517184936
],
[
158.936170212766,
368.94844517184936
],
[
165.936170212766,
337.94844517184936
],
[
347.936170212766,
335.94844517184936
],
[
349.936170212766,
369.94844517184936
],
[
391.936170212766,
373.94844517184936
],
[
403.936170212766,
335.94844517184936
],
[
425.936170212766,
332.94844517184936
],
[
421.936170212766,
281.94844517184936
],
[
428.936170212766,
252.94844517184944
],
[
428.936170212766,
236.94844517184944
],
[
409.936170212766,
220.94844517184944
],
[
409.936170212766,
150.94844517184944
],
[
430.936170212766,
143.94844517184944
],
[
433.936170212766,
112.94844517184944
],
[
431.936170212766,
96.94844517184944
],
[
408.936170212766,
90.94844517184944
],
[
395.936170212766,
50.94844517184944
],
[
338.936170212766,
25.948445171849443
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "bus",
"points": [
[
88.93617021276599,
115.56382978723406
],
[
0.9361702127659877,
96.56382978723406
],
[
0.0,
251.968085106388
],
[
0.9361702127659877,
265.56382978723406
],
[
27.936170212765987,
265.56382978723406
],
[
29.936170212765987,
283.56382978723406
],
[
63.93617021276599,
281.56382978723406
],
[
89.93617021276599,
252.56382978723406
],
[
100.93617021276599,
183.56382978723406
],
[
108.93617021276599,
145.56382978723406
],
[
81.93617021276599,
151.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
413.936170212766,
168.56382978723406
],
[
497.936170212766,
168.56382978723406
],
[
497.936170212766,
256.56382978723406
],
[
431.936170212766,
258.56382978723406
],
[
430.936170212766,
236.56382978723406
],
[
408.936170212766,
218.56382978723406
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "2011_000025.jpg",
"imageData": null,
"imageHeight": 375,
"imageWidth": 500
}
================================================
FILE: examples/semantic_segmentation/data_dataset_voc/class_names.txt
================================================
_background_
aeroplane
bicycle
bird
boat
bottle
bus
car
cat
chair
cow
diningtable
dog
horse
motorbike
person
potted plant
sheep
sofa
train
tv/monitor
================================================
FILE: examples/semantic_segmentation/labels.txt
================================================
__ignore__
_background_
aeroplane
bicycle
bird
boat
bottle
bus
car
cat
chair
cow
diningtable
dog
horse
motorbike
person
potted plant
sheep
sofa
train
tv/monitor
================================================
FILE: examples/tutorial/README.md
================================================
# Tutorial (Single Image Example)
## Annotation
```bash
labelme apc2016_obj3.jpg
```
## Visualization
To view the json file quickly, you can use utility script:
```bash
./draw_json.py apc2016_obj3.json
```
## Convert to Dataset
To convert the json to set of image and label, you can run following:
```bash
./export_json.py apc2016_obj3.json
```
It generates standard files from the JSON file.
- [img.png](apc2016_obj3/img.png): Image file.
- [label.png](apc2016_obj3/label.png): uint8 label file.
- [label_viz.png](apc2016_obj3/label_viz.png): Visualization of `label.png`.
- [label_names.txt](apc2016_obj3/label_names.txt): Label names for values in `label.png`.
## How to load label PNG file?
Note that loading `label.png` is a bit difficult
(`scipy.misc.imread`, `skimage.io.imread` may not work correctly),
and please use `PIL.Image.open` to avoid unexpected behavior:
```python
# see load_label_png.py also.
>>> import numpy as np
>>> import PIL.Image
>>> label_png = 'apc2016_obj3/label.png'
>>> lbl = np.asarray(PIL.Image.open(label_png))
>>> print(lbl.dtype)
dtype('uint8')
>>> np.unique(lbl)
array([0, 1, 2, 3], dtype=uint8)
>>> lbl.shape
(907, 1210)
```
Also, you can see the label PNG file by:
```bash
./draw_label_png.py apc2016_obj3/label.png
```
================================================
FILE: examples/tutorial/apc2016_obj3/label_names.txt
================================================
_background_
highland_6539_self_stick_notes
kong_air_dog_squeakair_tennis_ball
mead_index_cards
shelf
================================================
FILE: examples/tutorial/apc2016_obj3.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "shelf",
"points": [
[
7.942307692307736,
80.76150251617551
],
[
171.94230769230774,
714.7615025161755
],
[
968.9423076923077,
733.7615025161755
],
[
1181.9423076923076,
110.76150251617551
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "highland_6539_self_stick_notes",
"points": [
[
430.16339869281046,
516.2450980392157
],
[
390.9477124183006,
606.4411764705883
],
[
398.7908496732026,
697.2908496732026
],
[
522.3202614379085,
711.6699346405229
],
[
762.18954248366,
721.4738562091503
],
[
777.2222222222222,
634.5457516339869
],
[
761.5359477124183,
537.8137254901961
],
[
634.0849673202614,
518.8594771241831
],
[
573.3006535947712,
512.9771241830066
],
[
534.0849673202614,
513.6307189542484
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "mead_index_cards",
"points": [
[
447.156862745098,
394.6764705882353
],
[
410.55555555555554,
480.95098039215685
],
[
415.1307189542483,
522.781045751634
],
[
422.9738562091503,
522.781045751634
],
[
427.5490196078431,
515.5915032679738
],
[
570.032679738562,
512.3235294117648
],
[
732.1241830065359,
528.0098039215686
],
[
733.4313725490196,
492.71568627450984
],
[
724.9346405228757,
407.0947712418301
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "kong_air_dog_squeakair_tennis_ball",
"points": [
[
419.05228758169926,
266.5718954248366
],
[
343.235294117647,
303.1732026143791
],
[
511.2091503267974,
318.859477124183
],
[
519.7058823529411,
334.5457516339869
],
[
533.4313725490196,
339.7745098039216
],
[
551.7320261437908,
349.57843137254906
],
[
573.3006535947712,
354.15359477124184
],
[
592.2549019607843,
353.5
],
[
604.0196078431372,
345.0032679738562
],
[
613.1699346405228,
341.7352941176471
],
[
687.0261437908497,
365.91830065359477
],
[
696.1764705882352,
358.07516339869284
],
[
727.5490196078431,
329.3169934640523
],
[
677.2222222222222,
306.44117647058823
],
[
651.078431372549,
273.76143790849676
],
[
632.1241830065359,
272.4542483660131
],
[
612.516339869281,
248.27124183006538
],
[
596.8300653594771,
241.08169934640526
],
[
577.2222222222222,
234.54575163398695
],
[
563.4967320261437,
234.54575163398695
],
[
545.1960784313725,
235.19934640522877
],
[
534.7385620915032,
242.3888888888889
],
[
526.2418300653594,
247.61764705882354
],
[
513.1699346405228,
258.7287581699347
],
[
507.28758169934633,
266.5718954248366
],
[
502.0588235294117,
272.4542483660131
],
[
496.17647058823525,
273.1078431372549
],
[
474.6078431372549,
273.1078431372549
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "apc2016_obj3.json",
"imageData": "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",
"imageHeight": 907,
"imageWidth": 1210
}
================================================
FILE: examples/tutorial/draw_json.py
================================================
#!/usr/bin/env python
import argparse
import imgviz
import matplotlib.pyplot as plt
from labelme import utils
from labelme._label_file import LabelFile
def main():
parser = argparse.ArgumentParser()
parser.add_argument("json_file")
args = parser.parse_args()
label_file = LabelFile(args.json_file)
img = utils.img_data_to_arr(label_file.imageData)
label_name_to_value = {"_background_": 0}
for shape in sorted(label_file.shapes, key=lambda x: x["label"]):
label_name = shape["label"]
if label_name in label_name_to_value:
label_value = label_name_to_value[label_name]
else:
label_value = len(label_name_to_value)
label_name_to_value[label_name] = label_value
lbl, _ = utils.shapes_to_label(img.shape, label_file.shapes, label_name_to_value)
label_names: list[str] = [""] * (max(label_name_to_value.values()) + 1)
for name, value in label_name_to_value.items():
label_names[value] = name
lbl_viz = imgviz.label2rgb(
lbl,
imgviz.asgray(img),
label_names=label_names,
font_size=30,
loc="rb",
)
plt.subplot(121)
plt.imshow(img)
plt.subplot(122)
plt.imshow(lbl_viz)
plt.show()
if __name__ == "__main__":
main()
================================================
FILE: examples/tutorial/draw_label_png.py
================================================
#!/usr/bin/env python
import argparse
import os
import imgviz
import matplotlib.pyplot as plt
import numpy as np
from loguru import logger
def main():
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
parser.add_argument("label_png", help="label PNG file")
parser.add_argument(
"--labels",
help="labels list (comma separated text or file)",
default=None,
)
parser.add_argument("--image", help="image file", default=None)
args = parser.parse_args()
if args.labels is not None:
if os.path.exists(args.labels):
with open(args.labels) as f:
label_names = [label.strip() for label in f]
else:
label_names = args.labels.split(",")
else:
label_names = None
if args.image is not None:
image = imgviz.io.imread(args.image)
else:
image = None
label = imgviz.io.imread(args.label_png)
label = label.astype(np.int32)
label[label == 255] = -1
unique_label_values = np.unique(label)
logger.info(f"Label image shape: {label.shape}")
logger.info(f"Label values: {unique_label_values.tolist()}")
if label_names is not None:
logger.info(
"Label names: {}".format(
[
f"{label_value}:{label_names[label_value]}"
for label_value in unique_label_values
]
)
)
if args.image:
num_cols = 2
else:
num_cols = 1
plt.figure(figsize=(num_cols * 6, 5))
plt.subplot(1, num_cols, 1)
plt.title(args.label_png)
label_viz = imgviz.label2rgb(
label=label, label_names=label_names, font_size=label.shape[1] // 30
)
plt.imshow(label_viz)
if image is not None:
plt.subplot(1, num_cols, 2)
label_viz_with_overlay = imgviz.label2rgb(
label=label,
image=image,
label_names=label_names,
font_size=label.shape[1] // 30,
)
plt.title(f"{args.label_png}\n{args.image}")
plt.imshow(label_viz_with_overlay)
plt.tight_layout()
plt.show()
if __name__ == "__main__":
main()
================================================
FILE: examples/tutorial/export_json.py
================================================
#!/usr/bin/env python
import argparse
import os
import os.path as osp
import imgviz
import numpy as np
import PIL.Image
from loguru import logger
from numpy.typing import NDArray
from labelme import utils
from labelme._label_file import LabelFile
def main():
parser = argparse.ArgumentParser()
parser.add_argument("json_file")
parser.add_argument("-o", "--out", default=None)
args = parser.parse_args()
json_file = args.json_file
if args.out is None:
out_dir = osp.splitext(osp.basename(json_file))[0]
out_dir = osp.join(osp.dirname(json_file), out_dir)
else:
out_dir = args.out
os.makedirs(out_dir, exist_ok=True)
label_file: LabelFile = LabelFile(filename=json_file)
image: NDArray[np.uint8] = utils.img_data_to_arr(label_file.imageData)
label_name_to_value: dict[str, int] = {"_background_": 0}
for shape in sorted(label_file.shapes, key=lambda x: x["label"]):
label_name = shape["label"]
if label_name in label_name_to_value:
label_value = label_name_to_value[label_name]
else:
label_value = len(label_name_to_value)
label_name_to_value[label_name] = label_value
lbl, _ = utils.shapes_to_label(image.shape, label_file.shapes, label_name_to_value)
label_names: list[str] = [""] * (max(label_name_to_value.values()) + 1)
for name, value in label_name_to_value.items():
label_names[value] = name
lbl_viz = imgviz.label2rgb(
lbl, imgviz.asgray(image), label_names=label_names, loc="rb"
)
PIL.Image.fromarray(image).save(osp.join(out_dir, "img.png"))
utils.lblsave(osp.join(out_dir, "label.png"), lbl)
PIL.Image.fromarray(lbl_viz).save(osp.join(out_dir, "label_viz.png"))
with open(osp.join(out_dir, "label_names.txt"), "w") as f:
for lbl_name in label_names:
f.write(f"{lbl_name}\n")
logger.info(f"Saved to: {out_dir}")
if __name__ == "__main__":
main()
================================================
FILE: examples/tutorial/load_label_png.py
================================================
#!/usr/bin/env python
import os.path as osp
import numpy as np
import PIL.Image
here = osp.dirname(osp.abspath(__file__))
def main():
label_png = osp.join(here, "apc2016_obj3/label.png")
print("Loading:", label_png)
print()
lbl = np.asarray(PIL.Image.open(label_png))
labels = np.unique(lbl)
label_names_txt = osp.join(here, "apc2016_obj3/label_names.txt")
label_names = [name.strip() for name in open(label_names_txt)]
print("# of labels:", len(labels))
print("# of label_names:", len(label_names))
if len(labels) != len(label_names):
print("Number of unique labels and label_names must be same.")
quit(1)
print()
print("label: label_name")
for label, label_name in zip(labels, label_names):
print(f"{label}: {label_name}")
if __name__ == "__main__":
main()
================================================
FILE: examples/video_annotation/README.md
================================================
# Video Annotation Example
## Annotation
```bash
labelme data_annotated --labels labels.txt --keep-prev --config '{shift_auto_shape_color: -2}'
```
*Fig 1. Video annotation example. A frame (left), The next frame (right).*
*Fig 2. Visualization of video semantic segmentation.*
## How to Convert a Video File to Images for Annotation?
```bash
pip install video-cli
video-toimg your_video.mp4 # this creates your_video/ directory
ls your_video/
labelme your_video/
```
================================================
FILE: examples/video_annotation/data_annotated/00000100.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "track",
"points": [
[
634.0,
203.25925925925924
],
[
604.0,
274.25925925925924
],
[
603.0,
339.25925925925924
],
[
622.0,
362.25925925925924
],
[
639.0,
362.25925925925924
],
[
649.0,
353.25925925925924
],
[
682.0,
382.25925925925924
],
[
733.0,
389.25925925925924
],
[
748.0,
363.25925925925924
],
[
827.0,
358.25925925925924
],
[
829.0,
249.25925925925924
],
[
800.0,
193.25925925925924
],
[
775.0,
184.25925925925924
],
[
740.0,
198.25925925925924
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "track",
"points": [
[
860.0,
190.0
],
[
997.0,
186.0
],
[
998.0,
305.0
],
[
924.0,
320.0
],
[
905.0,
352.0
],
[
877.0,
353.0
],
[
869.0,
245.0
],
[
879.0,
222.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
924.0,
321.0
],
[
905.0,
352.0
],
[
909.0,
388.0
],
[
936.0,
404.0
],
[
959.0,
411.0
],
[
966.0,
431.0
],
[
1000.0,
432.0
],
[
1000.0,
306.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "00000100.jpg",
"imageData": null,
"imageHeight": 563,
"imageWidth": 1000
}
================================================
FILE: examples/video_annotation/data_annotated/00000101.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "track",
"points": [
[
614.7407407407408,
203.2592592592593
],
[
584.7407407407408,
274.2592592592593
],
[
583.7407407407408,
339.2592592592593
],
[
602.7407407407408,
362.2592592592593
],
[
619.7407407407408,
362.2592592592593
],
[
629.7407407407408,
353.2592592592593
],
[
662.7407407407408,
382.2592592592593
],
[
713.7407407407408,
389.2592592592593
],
[
728.7407407407408,
363.2592592592593
],
[
827.7407407407408,
357.2592592592593
],
[
825.7407407407408,
248.2592592592593
],
[
801.7407407407408,
199.2592592592593
],
[
757.7407407407408,
193.2592592592593
],
[
720.7407407407408,
198.2592592592593
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "track",
"points": [
[
860.0,
190.0
],
[
997.0,
186.0
],
[
998.0,
305.0
],
[
924.0,
320.0
],
[
905.0,
352.0
],
[
877.0,
353.0
],
[
869.0,
245.0
],
[
879.0,
222.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
924.0,
321.0
],
[
905.0,
352.0
],
[
909.0,
388.0
],
[
936.0,
404.0
],
[
959.0,
411.0
],
[
966.0,
431.0
],
[
1000.0,
432.0
],
[
1000.0,
306.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "00000101.jpg",
"imageData": null,
"imageHeight": 563,
"imageWidth": 1000
}
================================================
FILE: examples/video_annotation/data_annotated/00000102.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "track",
"points": [
[
591.5185185185185,
202.51851851851853
],
[
561.5185185185185,
273.51851851851853
],
[
560.5185185185185,
338.51851851851853
],
[
579.5185185185185,
361.51851851851853
],
[
596.5185185185185,
361.51851851851853
],
[
606.5185185185185,
352.51851851851853
],
[
639.5185185185185,
381.51851851851853
],
[
690.5185185185185,
388.51851851851853
],
[
705.5185185185185,
362.51851851851853
],
[
825.5185185185185,
356.51851851851853
],
[
821.5185185185185,
241.51851851851853
],
[
800.5185185185185,
197.51851851851853
],
[
734.5185185185185,
192.51851851851853
],
[
697.5185185185185,
197.51851851851853
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "track",
"points": [
[
860.0,
190.0
],
[
997.0,
186.0
],
[
998.0,
305.0
],
[
924.0,
320.0
],
[
905.0,
352.0
],
[
877.0,
353.0
],
[
869.0,
245.0
],
[
879.0,
222.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
924.0,
321.0
],
[
905.0,
352.0
],
[
909.0,
388.0
],
[
936.0,
404.0
],
[
959.0,
411.0
],
[
966.0,
431.0
],
[
1000.0,
432.0
],
[
1000.0,
306.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "00000102.jpg",
"imageData": null,
"imageHeight": 563,
"imageWidth": 1000
}
================================================
FILE: examples/video_annotation/data_annotated/00000103.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "track",
"points": [
[
573.0,
202.55555555555554
],
[
543.0,
273.55555555555554
],
[
542.0,
338.55555555555554
],
[
561.0,
361.55555555555554
],
[
578.0,
361.55555555555554
],
[
588.0,
352.55555555555554
],
[
621.0,
381.55555555555554
],
[
672.0,
388.55555555555554
],
[
687.0,
362.55555555555554
],
[
829.0,
349.55555555555554
],
[
821.0,
231.55555555555554
],
[
801.0,
194.55555555555554
],
[
716.0,
192.55555555555554
],
[
679.0,
197.55555555555554
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "track",
"points": [
[
860.0,
190.0
],
[
997.0,
186.0
],
[
998.0,
305.0
],
[
924.0,
320.0
],
[
905.0,
352.0
],
[
877.0,
353.0
],
[
869.0,
245.0
],
[
879.0,
222.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
924.0,
321.0
],
[
905.0,
352.0
],
[
909.0,
388.0
],
[
936.0,
404.0
],
[
959.0,
411.0
],
[
966.0,
431.0
],
[
1000.0,
432.0
],
[
1000.0,
306.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "00000103.jpg",
"imageData": null,
"imageHeight": 563,
"imageWidth": 1000
}
================================================
FILE: examples/video_annotation/data_annotated/00000104.json
================================================
{
"version": "4.0.0",
"flags": {},
"shapes": [
{
"label": "track",
"points": [
[
555.2592592592594,
200.25925925925924
],
[
527.2592592592594,
276.25925925925924
],
[
523.2592592592594,
341.25925925925924
],
[
527.2592592592594,
360.25925925925924
],
[
562.2592592592594,
364.25925925925924
],
[
572.2592592592594,
355.25925925925924
],
[
605.2592592592594,
384.25925925925924
],
[
656.2592592592594,
391.25925925925924
],
[
671.2592592592594,
365.25925925925924
],
[
824.2592592592594,
353.25925925925924
],
[
825.2592592592594,
237.25925925925924
],
[
800.2592592592594,
201.25925925925924
],
[
700.2592592592594,
195.25925925925924
],
[
663.2592592592594,
200.25925925925924
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "track",
"points": [
[
860.0,
190.0
],
[
997.0,
186.0
],
[
998.0,
305.0
],
[
924.0,
320.0
],
[
905.0,
352.0
],
[
874.0,
354.0
],
[
869.0,
245.0
],
[
879.0,
222.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
},
{
"label": "car",
"points": [
[
924.0,
321.0
],
[
905.0,
352.0
],
[
909.0,
388.0
],
[
936.0,
404.0
],
[
959.0,
411.0
],
[
966.0,
431.0
],
[
1000.0,
432.0
],
[
1000.0,
306.0
]
],
"group_id": null,
"shape_type": "polygon",
"flags": {}
}
],
"imagePath": "00000104.jpg",
"imageData": null,
"imageHeight": 563,
"imageWidth": 1000
}
================================================
FILE: examples/video_annotation/data_dataset_voc/class_names.txt
================================================
_background_
car
track
================================================
FILE: examples/video_annotation/labels.txt
================================================
__ignore__
_background_
car
track
================================================
FILE: labelme/__init__.py
================================================
import importlib.metadata
__appname__ = "Labelme"
# Semantic Versioning 2.0.0: https://semver.org/
# 1. MAJOR version when you make incompatible API changes;
# 2. MINOR version when you add functionality in a backwards-compatible manner;
# 3. PATCH version when you make backwards-compatible bug fixes.
# e.g., 1.0.0a0, 1.0.0a1, 1.0.0b0, 1.0.0rc0, 1.0.0, 1.0.0.post0
__version__ = importlib.metadata.version("labelme")
# XXX: has to be imported before PyQt5 to load dlls in order on Windows
# https://github.com/wkentaro/labelme/issues/1564
import onnxruntime
from labelme import utils
from labelme._label_file import LabelFile
================================================
FILE: labelme/__main__.py
================================================
import argparse
import codecs
import contextlib
import io
import os
import os.path as osp
import sys
import traceback
import warnings
from pathlib import Path
from typing import AnyStr
import yaml
from loguru import logger
from PyQt5 import QtCore
from PyQt5 import QtWidgets
from labelme import __appname__
from labelme import __version__
from labelme.app import MainWindow
from labelme.config import get_user_config_file
from labelme.utils import newIcon
class _LoggerIO(io.StringIO):
def write(self, s: AnyStr) -> int:
assert isinstance(s, str)
if stripped_s := s.strip():
logger.debug(stripped_s)
return len(s)
def flush(self) -> None:
pass
def writable(self) -> bool:
return True
def readable(self) -> bool:
return False
def seekable(self) -> bool:
return False
@property
def closed(self) -> bool:
return False
def _setup_loguru(logger_level: str) -> None:
try:
logger.remove(handler_id=0)
except ValueError:
pass
if sys.stderr:
logger.add(sys.stderr, level=logger_level)
cache_dir: str
if os.name == "nt":
cache_dir = os.path.join(os.environ["LOCALAPPDATA"], "labelme")
else:
cache_dir = os.path.expanduser("~/.cache/labelme")
os.makedirs(cache_dir, exist_ok=True)
log_file = os.path.join(cache_dir, "labelme.log")
logger.add(
log_file,
colorize=True,
level="DEBUG",
rotation="10 MB",
retention="30 days",
compression="gz",
enqueue=True,
backtrace=True,
diagnose=True,
)
def _handle_exception(exc_type, exc_value, exc_traceback):
if issubclass(exc_type, KeyboardInterrupt):
sys.__excepthook__(exc_type, exc_value, exc_traceback)
sys.exit(0)
traceback_str: str = "".join(
traceback.format_exception(exc_type, exc_value, exc_traceback)
)
logger.critical(traceback_str)
traceback_html: str = traceback_str.replace("\n", "").replace(" ", " ") QtWidgets.QMessageBox.critical( None, "Error", f"An unexpected error occurred. The application will close.
Please report issues following the Troubleshoot.
{traceback_html}", # noqa: E501 ) if app := QtWidgets.QApplication.instance(): app.quit() sys.exit(1) def main(): parser = argparse.ArgumentParser() parser.add_argument("--version", "-V", action="store_true", help="show version") parser.add_argument("--reset-config", action="store_true", help="reset qt config") parser.add_argument( "--logger-level", default="debug", choices=["debug", "info", "warning", "fatal", "error"], help="logger level", ) parser.add_argument("path", nargs="?", help="image file, label file, or directory") parser.add_argument( "--output", help="output directory for saving annotation JSON files", ) default_config_file = get_user_config_file() parser.add_argument( "--config", dest="config", help=f"config file or yaml-format string (default: {default_config_file})", default=default_config_file, ) # config for the gui parser.add_argument( "--nodata", dest="_deprecated_nodata", action="store_true", help=argparse.SUPPRESS, default=argparse.SUPPRESS, ) parser.add_argument( "--with-image-data", dest="with_image_data", action="store_true", help="store image data in JSON file", default=argparse.SUPPRESS, ) parser.add_argument( "--no-auto-save", dest="auto_save", action="store_false", help="disable auto save", default=argparse.SUPPRESS, ) parser.add_argument( "--autosave", dest="_deprecated_autosave", action="store_true", help=argparse.SUPPRESS, default=argparse.SUPPRESS, ) parser.add_argument( "--no-sort-labels", "--nosortlabels", # deprecated dest="sort_labels", action="store_false", help="stop sorting labels", default=argparse.SUPPRESS, ) parser.add_argument( "--flags", help="comma separated list of flags OR file containing flags", default=argparse.SUPPRESS, ) parser.add_argument( "--label-flags", "--labelflags", # deprecated dest="label_flags", help=r"yaml string of label specific flags OR file containing json " r"string of label specific flags (ex. {person-\d+: [male, tall], " r"dog-\d+: [black, brown, white], .*: [occluded]})", # NOQA default=argparse.SUPPRESS, ) parser.add_argument( "--labels", help="comma separated list of labels OR file containing labels", default=argparse.SUPPRESS, ) parser.add_argument( "--validate-label", "--validatelabel", # deprecated dest="validate_label", choices=["exact"], help="label validation types", default=argparse.SUPPRESS, ) parser.add_argument( "--keep-prev", action="store_true", help="keep annotation of previous frame", default=argparse.SUPPRESS, ) parser.add_argument( "--epsilon", type=float, help="epsilon to find nearest vertex on canvas", default=argparse.SUPPRESS, ) args = parser.parse_args() if hasattr(args, "_deprecated_nodata"): warnings.warn( "--nodata is deprecated and will be removed in a future version. " "Image data is no longer stored by default. " "Use --with-image-data to store it.", FutureWarning, stacklevel=1, ) del args._deprecated_nodata if hasattr(args, "_deprecated_autosave"): warnings.warn( "--autosave is deprecated and will be removed in a future version. " "Auto save is now enabled by default. Use --no-autosave to disable it.", FutureWarning, stacklevel=1, ) del args._deprecated_autosave if args.version: print(f"{__appname__} {__version__}") sys.exit(0) _setup_loguru(logger_level=args.logger_level.upper()) logger.info("Starting {} {}", __appname__, __version__) sys.excepthook = _handle_exception if hasattr(args, "flags"): if os.path.isfile(args.flags): with codecs.open(args.flags, "r", encoding="utf-8") as f: args.flags = [line.strip() for line in f if line.strip()] else: args.flags = [line for line in args.flags.split(",") if line] if hasattr(args, "labels"): if os.path.isfile(args.labels): with codecs.open(args.labels, "r", encoding="utf-8") as f: args.labels = [line.strip() for line in f if line.strip()] else: args.labels = [line for line in args.labels.split(",") if line] if hasattr(args, "label_flags"): if os.path.isfile(args.label_flags): with codecs.open(args.label_flags, "r", encoding="utf-8") as f: args.label_flags = yaml.safe_load(f) else: args.label_flags = yaml.safe_load(args.label_flags) config_from_args = args.__dict__ config_from_args.pop("version") reset_config = config_from_args.pop("reset_config") filename = config_from_args.pop("path") output = config_from_args.pop("output") config_overrides: dict config_file: Path | None config_str: str = config_from_args.pop("config") if isinstance(config_loaded := yaml.safe_load(config_str), dict): config_overrides = config_loaded config_file = None else: config_overrides = {} config_file = Path(config_str) if not config_file.is_file(): logger.error( "Config file does not exist: {!r}", str(config_file.absolute()) ) sys.exit(1) del config_str config_overrides.update(config_from_args) output_dir = None if output is not None: if output.endswith(".json"): parser.error( f"--output expects a directory path, but '{output}' looks like a file." " Remove the .json extension or provide a directory path." ) output_dir = output translator = QtCore.QTranslator() translator.load( QtCore.QLocale.system().name(), f"{osp.dirname(osp.abspath(__file__))}/translate", ) app = QtWidgets.QApplication(sys.argv) app.setStyle("Fusion") # for consistent appearance across platforms # Force light mode to avoid dark mode UI issues (e.g., invisible icons) app.setPalette(QtWidgets.QStyleFactory.create("Fusion").standardPalette()) app.setApplicationName(__appname__) app.setWindowIcon(newIcon("icon")) app.installTranslator(translator) win = MainWindow( config_file=config_file, config_overrides=config_overrides, filename=filename, output_dir=output_dir, ) if reset_config: logger.info(f"Resetting Qt config: {win.settings.fileName()}") win.settings.clear() sys.exit(0) with contextlib.redirect_stderr(new_target=_LoggerIO()): win.show() win.raise_() sys.exit(app.exec_()) # this main block is required to generate executable by pyinstaller if __name__ == "__main__": main() ================================================ FILE: labelme/_automation/__init__.py ================================================ from ._osam_session import OsamSession ================================================ FILE: labelme/_automation/_osam_session.py ================================================ from __future__ import annotations import collections import numpy as np import osam from loguru import logger from numpy.typing import NDArray class OsamSession: _model_name: str _model: osam.types.Model | None _embedding_cache: collections.deque[tuple[str, osam.types.ImageEmbedding]] def __init__( self, model_name: str = "sam2:latest", embedding_cache_size: int = 3, ) -> None: logger.debug("Initializing OsamSession with model_name={!r}", model_name) self._model_name = model_name self._model = None self._embedding_cache = collections.deque(maxlen=embedding_cache_size) logger.debug("Initialized OsamSession with model_name={!r}", model_name) @property def model_name(self) -> str: return self._model_name def run( self, image: NDArray[np.uint8], image_id: str, points: NDArray[np.floating] | None = None, point_labels: NDArray[np.intp] | None = None, texts: list[str] | None = None, ) -> osam.types.GenerateResponse: image_embedding: osam.types.ImageEmbedding | None try: image_embedding = self._get_or_compute_embedding( image=image, image_id=image_id ) except NotImplementedError: image_embedding = None prompt: osam.types.Prompt if points is not None and point_labels is not None: prompt = osam.types.Prompt( points=points, point_labels=point_labels, ) elif texts is not None: prompt = osam.types.Prompt( texts=texts, iou_threshold=1.0, score_threshold=0.01, max_annotations=1000, ) else: raise ValueError( "Either points and point_labels, or texts must be provided." ) model: osam.types.Model = self._get_or_load_model() return model.generate( request=osam.types.GenerateRequest( model=model.name, image=image, image_embedding=image_embedding, prompt=prompt, ) ) def _get_or_compute_embedding( self, image: NDArray[np.uint8], image_id: str ) -> osam.types.ImageEmbedding: for key, embedding in self._embedding_cache: if key == image_id: return embedding model: osam.types.Model = self._get_or_load_model() logger.debug("Computing embedding for cache_key={!r}", image_id) embedding: osam.types.ImageEmbedding = model.encode_image(image=image) self._embedding_cache.append((image_id, embedding)) logger.debug("Cached embedding for cache_key={!r}", image_id) return embedding def _get_or_load_model(self) -> osam.types.Model: if self._model is None: logger.debug("Loading model with name={!r}", self._model_name) self._model = osam.apis.get_model_type_by_name(self._model_name)() logger.debug("Loaded model with name={!r}", self._model_name) return self._model ================================================ FILE: labelme/_automation/bbox_from_text.py ================================================ import json import time from typing import Literal import numpy as np import osam from loguru import logger from numpy.typing import NDArray from PyQt5 import QtCore from labelme.shape import Shape from ._osam_session import OsamSession from .polygon_from_mask import compute_polygon_from_mask def get_bboxes_from_texts( session: OsamSession, image: np.ndarray, image_id: str, texts: list[str] ) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray | None]: logger.debug( f"Requesting with model={session.model_name!r}, " f"image={(image.shape, image.dtype)}, texts={texts!r}" ) t_start: float = time.time() response: osam.types.GenerateResponse = session.run( image=image, image_id=image_id, texts=texts, ) num_annotations: int = len(response.annotations) logger.debug( f"Response: num_annotations={num_annotations}, " f"elapsed_time={time.time() - t_start:.3f} [s]" ) boxes: NDArray[np.float32] = np.empty((num_annotations, 4), dtype=np.float32) scores: NDArray[np.float32] = np.empty((num_annotations,), dtype=np.float32) labels: NDArray[np.int32] = np.empty((num_annotations,), dtype=np.int32) for i, annotation in enumerate(response.annotations): if annotation.bounding_box is None: raise ValueError("Bounding box is missing in the annotation.") if annotation.text not in texts: raise ValueError( f"Unexpected text {annotation.text!r} found in the response." ) boxes[i] = [ annotation.bounding_box.xmin, annotation.bounding_box.ymin, annotation.bounding_box.xmax, annotation.bounding_box.ymax, ] scores[i] = annotation.score labels[i] = texts.index(annotation.text) masks: NDArray[np.bool_] | None = None if response.annotations and response.annotations[0].mask is not None: masks = np.array( [annotation.mask for annotation in response.annotations], dtype=np.bool_ ) return boxes, scores, labels, masks def nms_bboxes( boxes: np.ndarray, scores: np.ndarray, labels: np.ndarray, iou_threshold: float, score_threshold: float, max_num_detections: int, ) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]: if len(boxes) == 0: return boxes, scores, labels, np.empty((0,), dtype=np.int32) num_classes: int = max(labels) + 1 scores_of_all_classes: NDArray[np.float32] = np.zeros( (len(boxes), num_classes), dtype=np.float32 ) for i, (score, label) in enumerate(zip(scores, labels)): scores_of_all_classes[i, label] = score logger.debug( "Running NMS: iou_threshold={}, score_threshold={}, max_num_detections={}", iou_threshold, score_threshold, max_num_detections, ) logger.debug(f"Input: num_boxes={len(boxes)}") boxes, scores, labels, indices = osam.apis.non_maximum_suppression( boxes=boxes, scores=scores_of_all_classes, iou_threshold=iou_threshold, score_threshold=score_threshold, max_num_detections=max_num_detections, ) logger.debug(f"Output: num_boxes={len(boxes)}") return boxes, scores, labels, indices def get_shapes_from_bboxes( boxes: np.ndarray, scores: np.ndarray, labels: np.ndarray, texts: list[str], masks: NDArray[np.bool_] | None, shape_type: Literal["rectangle", "polygon", "mask"], ) -> list[Shape]: shapes: list[Shape] = [] for i, (box, score, label) in enumerate(zip(boxes, scores, labels)): text: str = texts[label] xmin, ymin, xmax, ymax = box points: list[list[float]] mask: NDArray[np.bool_] | None = None if shape_type == "rectangle": points = [[xmin, ymin], [xmax, ymax]] elif shape_type == "polygon": if masks is None: points = [ [xmin, ymin], [xmax, ymin], [xmax, ymax], [xmin, ymax], [xmin, ymin], ] else: points = compute_polygon_from_mask(mask=masks[i]).tolist() elif shape_type == "mask": xmin = int(xmin) ymin = int(ymin) xmax = int(xmax) ymax = int(ymax) points = [[xmin, ymin], [xmax, ymax]] if masks is None: mask = np.zeros((ymax - ymin, xmax - xmin), dtype=bool) else: mask = masks[i][ymin : ymax + 1, xmin : xmax + 1] else: raise ValueError(f"Unsupported shape_type: {shape_type!r}") shape = Shape( label=text, shape_type=shape_type, mask=mask, description=json.dumps(dict(score=score.item(), text=text)), ) for point in points: shape.addPoint(QtCore.QPointF(point[0], point[1])) shapes.append(shape) return shapes ================================================ FILE: labelme/_automation/polygon_from_mask.py ================================================ import imgviz import numpy as np import skimage from loguru import logger from numpy.typing import NDArray def _get_contour_length(contour: NDArray[np.float32]) -> float: contour_start: NDArray[np.float32] = contour contour_end: NDArray[np.float32] = np.r_[contour[1:], contour[0:1]] return np.linalg.norm(contour_end - contour_start, axis=1).sum() def compute_polygon_from_mask(mask: NDArray[np.bool_]) -> NDArray[np.float32]: contours: NDArray[np.float32] = skimage.measure.find_contours( np.pad(mask, pad_width=1) ) if len(contours) == 0: logger.warning("No contour found, so returning empty polygon.") return np.empty((0, 2), dtype=np.float32) contour: NDArray[np.float32] = max(contours, key=_get_contour_length) POLYGON_APPROX_TOLERANCE: float = 0.004 polygon: NDArray[np.float32] = skimage.measure.approximate_polygon( coords=contour, tolerance=np.ptp(contour, axis=0).max() * POLYGON_APPROX_TOLERANCE, ) polygon = np.clip(polygon, (0, 0), (mask.shape[0] - 1, mask.shape[1] - 1)) polygon = polygon[:-1] # drop last point that is duplicate of first point if 0: import PIL.Image image_pil = PIL.Image.fromarray(imgviz.gray2rgb(imgviz.bool2ubyte(mask))) imgviz.draw.line_(image_pil, yx=polygon, fill=(0, 255, 0)) for point in polygon: imgviz.draw.circle_(image_pil, center=point, diameter=10, fill=(0, 255, 0)) imgviz.io.imsave("contour.jpg", np.asarray(image_pil)) return polygon[:, ::-1] # yx -> xy ================================================ FILE: labelme/_label_file.py ================================================ from __future__ import annotations import base64 import contextlib import io import json import os.path as osp import time from pathlib import PureWindowsPath from typing import TypedDict import numpy as np import PIL.Image import tifffile from loguru import logger from numpy.typing import NDArray from labelme import __version__ from labelme import utils PIL.Image.MAX_IMAGE_PIXELS = None @contextlib.contextmanager def _open(name, mode): assert mode in ["r", "w"] encoding = "utf-8" yield open(name, mode, encoding=encoding) return class ShapeDict(TypedDict): label: str points: list[list[float]] shape_type: str flags: dict[str, bool] description: str group_id: int | None mask: NDArray[np.bool_] | None other_data: dict def _load_shape_json_obj(shape_json_obj: dict) -> ShapeDict: SHAPE_KEYS: set[str] = { "label", "points", "group_id", "shape_type", "flags", "description", "mask", } if "label" not in shape_json_obj: raise ValueError(f"label is required: {shape_json_obj}") if not isinstance(shape_json_obj["label"], str): raise TypeError(f"label must be str: {shape_json_obj['label']}") label: str = shape_json_obj["label"] if "points" not in shape_json_obj: raise ValueError(f"points is required: {shape_json_obj}") if not isinstance(shape_json_obj["points"], list): raise TypeError(f"points must be list: {shape_json_obj['points']}") if not shape_json_obj["points"]: raise ValueError(f"points must be non-empty: {shape_json_obj}") if not all( isinstance(point, list) and len(point) == 2 and all(isinstance(xy, int | float) for xy in point) for point in shape_json_obj["points"] ): raise ValueError(f"points must be list of [x, y]: {shape_json_obj['points']}") points: list[list[float]] = shape_json_obj["points"] if "shape_type" not in shape_json_obj: raise ValueError(f"shape_type is required: {shape_json_obj}") if not isinstance(shape_json_obj["shape_type"], str): raise TypeError(f"shape_type must be str: {shape_json_obj['shape_type']}") shape_type: str = shape_json_obj["shape_type"] flags: dict = {} if shape_json_obj.get("flags") is not None: if not isinstance(shape_json_obj["flags"], dict): raise TypeError(f"flags must be dict: {shape_json_obj['flags']}") if not all( isinstance(k, str) and isinstance(v, bool) for k, v in shape_json_obj["flags"].items() ): raise TypeError( f"flags must be dict of str to bool: {shape_json_obj['flags']}" ) flags = shape_json_obj["flags"] description: str = "" if shape_json_obj.get("description") is not None: if not isinstance(shape_json_obj["description"], str): raise TypeError(f"description must be str: {shape_json_obj['description']}") description = shape_json_obj["description"] group_id: int | None = None if shape_json_obj.get("group_id") is not None: if not isinstance(shape_json_obj["group_id"], int): raise TypeError(f"group_id must be int: {shape_json_obj['group_id']}") group_id = shape_json_obj["group_id"] mask: NDArray[np.bool_] | None = None if shape_json_obj.get("mask") is not None: if not isinstance(shape_json_obj["mask"], str): raise TypeError( f"mask must be base64-encoded PNG: {shape_json_obj['mask']}" ) mask = utils.img_b64_to_arr(shape_json_obj["mask"]).astype(bool) other_data = {k: v for k, v in shape_json_obj.items() if k not in SHAPE_KEYS} loaded: ShapeDict = ShapeDict( label=label, points=points, shape_type=shape_type, flags=flags, description=description, group_id=group_id, mask=mask, other_data=other_data, ) if set(loaded.keys()) != SHAPE_KEYS | {"other_data"}: raise RuntimeError( f"unexpected keys: {set(loaded.keys())} != {SHAPE_KEYS | {'other_data'}}" ) return loaded class LabelFileError(Exception): pass class LabelFile: shapes: list[ShapeDict] suffix = ".json" def __init__(self, filename: str | None = None): self.shapes: list[ShapeDict] = [] self.imagePath: str | None = None self.imageData: bytes | None = None if filename is not None: self.load(filename) self.filename: str | None = filename @staticmethod def load_image_file(filename): t0 = time.time() image_pil = _imread(filename=filename) oriented: PIL.Image.Image = utils.apply_exif_orientation(image_pil) ext = osp.splitext(filename)[1].lower() if oriented is image_pil and ext in (".jpg", ".jpeg", ".png"): # no encoding needed with open(filename, "rb") as f: imageData = f.read() else: with io.BytesIO() as f: format = "PNG" if "A" in oriented.mode else "JPEG" oriented.save(f, format=format, quality=95) f.seek(0) imageData = f.read() logger.debug( "Loaded image file: {!r} in {:.0f}ms", filename, (time.time() - t0) * 1000 ) return imageData def load(self, filename): keys = [ "version", "imageData", "imagePath", "shapes", # polygonal annotations "flags", # image level flags "imageHeight", "imageWidth", ] try: with _open(filename, "r") as f: data = json.load(f) # Normalize Windows-style backslash paths to POSIX forward slashes imagePath = PureWindowsPath(data["imagePath"]).as_posix() if data["imageData"] is not None: imageData = base64.b64decode(data["imageData"]) else: # relative path from label file to relative path from cwd imageData = self.load_image_file( osp.join(osp.dirname(filename), imagePath) ) flags = data.get("flags") or {} self._check_image_height_and_width( imageData, data.get("imageHeight"), data.get("imageWidth"), ) shapes: list[ShapeDict] = [ _load_shape_json_obj(shape_json_obj=s) for s in data["shapes"] ] except Exception as e: raise LabelFileError(e) otherData = {} for key, value in data.items(): if key not in keys: otherData[key] = value # Only replace data after everything is loaded. self.flags = flags self.shapes = shapes self.imagePath = imagePath self.imageData = imageData self.filename = filename self.otherData = otherData @staticmethod def _check_image_height_and_width(imageData, imageHeight, imageWidth): img_pil = utils.img_data_to_pil(imageData) actual_w, actual_h = img_pil.size if imageHeight is not None and actual_h != imageHeight: logger.error( "imageHeight does not match with imageData or imagePath, " "so getting imageHeight from actual image." ) imageHeight = actual_h if imageWidth is not None and actual_w != imageWidth: logger.error( "imageWidth does not match with imageData or imagePath, " "so getting imageWidth from actual image." ) imageWidth = actual_w return imageHeight, imageWidth def save( self, filename, shapes, imagePath, imageHeight, imageWidth, imageData=None, otherData=None, flags=None, ): if imageData is not None: imageHeight, imageWidth = self._check_image_height_and_width( imageData, imageHeight, imageWidth ) imageData = base64.b64encode(imageData).decode("utf-8") if otherData is None: otherData = {} if flags is None: flags = {} data = dict( version=__version__, flags=flags, shapes=shapes, imagePath=imagePath, imageData=imageData, imageHeight=imageHeight, imageWidth=imageWidth, ) for key, value in otherData.items(): assert key not in data data[key] = value try: with _open(filename, "w") as f: json.dump(data, f, ensure_ascii=False, indent=2) self.filename = filename except Exception as e: raise LabelFileError(e) @staticmethod def is_label_file(filename): return osp.splitext(filename)[1].lower() == LabelFile.suffix _DISPLAYABLE_MODES = {"1", "L", "P", "RGB", "RGBA", "LA", "PA"} def _imread(filename: str) -> PIL.Image.Image: ext: str = osp.splitext(filename)[1].lower() try: image_pil = PIL.Image.open(filename) if image_pil.mode not in _DISPLAYABLE_MODES: raise PIL.UnidentifiedImageError return image_pil except PIL.UnidentifiedImageError: if ext in (".tif", ".tiff"): return _imread_tiff(filename) raise def _imread_tiff(filename: str) -> PIL.Image.Image: img_arr: NDArray = tifffile.imread(filename) if img_arr.ndim == 2: img_arr_normalized = _normalize_to_uint8(img_arr) elif img_arr.ndim == 3: if img_arr.shape[2] >= 3: img_arr_normalized = np.stack( [_normalize_to_uint8(img_arr[:, :, i]) for i in range(3)], axis=2, ) else: img_arr_normalized = _normalize_to_uint8(img_arr[:, :, 0]) else: raise OSError(f"Unsupported image shape: {img_arr.shape}") return PIL.Image.fromarray(img_arr_normalized) def _normalize_to_uint8(arr: NDArray) -> NDArray[np.uint8]: arr = arr.astype(np.float64) min_val = np.nanmin(arr) max_val = np.nanmax(arr) if np.isnan(min_val) or np.isnan(max_val) or max_val - min_val == 0: return np.zeros(arr.shape, dtype=np.uint8) normalized = (arr - min_val) / (max_val - min_val) * 255 return np.clip(normalized, 0, 255).astype(np.uint8) ================================================ FILE: labelme/app.py ================================================ from __future__ import annotations import enum import functools import html import math import os import os.path as osp import platform import re import subprocess import time import webbrowser from collections.abc import Callable from pathlib import Path from typing import Literal from typing import NamedTuple import imgviz import natsort import numpy as np import osam from loguru import logger from numpy.typing import NDArray from PyQt5 import QtCore from PyQt5 import QtGui from PyQt5 import QtWidgets from PyQt5.QtCore import Qt from PyQt5.QtWidgets import QMessageBox from labelme import __appname__ from labelme import __version__ from labelme._automation import bbox_from_text from labelme._automation._osam_session import OsamSession from labelme._label_file import LabelFile from labelme._label_file import LabelFileError from labelme._label_file import ShapeDict from labelme.config import load_config from labelme.shape import Shape from labelme.widgets import AiAssistedAnnotationWidget from labelme.widgets import AiTextToAnnotationWidget from labelme.widgets import BrightnessContrastDialog from labelme.widgets import Canvas from labelme.widgets import FileDialogPreview from labelme.widgets import LabelDialog from labelme.widgets import LabelListWidget from labelme.widgets import LabelListWidgetItem from labelme.widgets import StatusStats from labelme.widgets import ToolBar from labelme.widgets import UniqueLabelQListWidget from labelme.widgets import ZoomWidget from labelme.widgets import download_ai_model from . import utils # handle high-dpi scaling issue # https://leomoon.com/journal/python/high-dpi-scaling-in-pyqt5 if hasattr(QtCore.Qt, "AA_EnableHighDpiScaling"): QtWidgets.QApplication.setAttribute(QtCore.Qt.AA_EnableHighDpiScaling, True) if hasattr(QtCore.Qt, "AA_UseHighDpiPixmaps"): QtWidgets.QApplication.setAttribute(QtCore.Qt.AA_UseHighDpiPixmaps, True) LABEL_COLORMAP: NDArray[np.uint8] = imgviz.label_colormap() class _ZoomMode(enum.Enum): FIT_WINDOW = enum.auto() FIT_WIDTH = enum.auto() MANUAL_ZOOM = enum.auto() _AI_TEXT_TO_ANNOTATION_CREATE_MODE_TO_SHAPE_TYPE: dict[ str, Literal["mask", "polygon", "rectangle"] ] = { "ai_mask": "mask", "ai_polygon": "polygon", "polygon": "polygon", "rectangle": "rectangle", } class _StatusBarWidgets(NamedTuple): message: QtWidgets.QLabel stats: StatusStats class _CanvasWidgets(NamedTuple): canvas: Canvas zoom_widget: ZoomWidget scroll_bars: dict[Qt.Orientation, QtWidgets.QScrollBar] class _DockWidgets(NamedTuple): flag_dock: QtWidgets.QDockWidget flag_list: QtWidgets.QListWidget shape_dock: QtWidgets.QDockWidget label_list: LabelListWidget label_dock: QtWidgets.QDockWidget unique_label_list: UniqueLabelQListWidget file_dock: QtWidgets.QDockWidget file_search: QtWidgets.QLineEdit file_list: QtWidgets.QListWidget class _Actions(NamedTuple): about: QtWidgets.QAction save: QtWidgets.QAction save_as: QtWidgets.QAction save_auto: QtWidgets.QAction save_with_image_data: QtWidgets.QAction change_output_dir: QtWidgets.QAction open: QtWidgets.QAction close: QtWidgets.QAction delete_file: QtWidgets.QAction toggle_keep_prev_mode: QtWidgets.QAction toggle_keep_prev_brightness_contrast: QtWidgets.QAction delete: QtWidgets.QAction edit: QtWidgets.QAction duplicate: QtWidgets.QAction copy: QtWidgets.QAction paste: QtWidgets.QAction undo_last_point: QtWidgets.QAction undo: QtWidgets.QAction remove_point: QtWidgets.QAction create_mode: QtWidgets.QAction edit_mode: QtWidgets.QAction create_rectangle_mode: QtWidgets.QAction create_circle_mode: QtWidgets.QAction create_line_mode: QtWidgets.QAction create_point_mode: QtWidgets.QAction create_line_strip_mode: QtWidgets.QAction create_ai_polygon_mode: QtWidgets.QAction create_ai_mask_mode: QtWidgets.QAction open_next_img: QtWidgets.QAction open_prev_img: QtWidgets.QAction keep_prev_scale: QtWidgets.QAction fit_window: QtWidgets.QAction fit_width: QtWidgets.QAction brightness_contrast: QtWidgets.QAction zoom_in: QtWidgets.QAction zoom_out: QtWidgets.QAction zoom_org: QtWidgets.QAction reset_layout: QtWidgets.QAction fill_drawing: QtWidgets.QAction hide_all: QtWidgets.QAction show_all: QtWidgets.QAction toggle_all: QtWidgets.QAction open_dir: QtWidgets.QAction zoom_widget_action: QtWidgets.QWidgetAction draw: list[tuple[str, QtWidgets.QAction]] zoom: tuple[ZoomWidget | QtWidgets.QAction, ...] on_load_active: tuple[QtWidgets.QAction, ...] on_shapes_present: tuple[QtWidgets.QAction, ...] context_menu: tuple[QtWidgets.QAction, ...] edit_menu: tuple[QtWidgets.QAction | None, ...] class _Menus(NamedTuple): file: QtWidgets.QMenu edit: QtWidgets.QMenu view: QtWidgets.QMenu help: QtWidgets.QMenu recent_files: QtWidgets.QMenu label_list: QtWidgets.QMenu class MainWindow(QtWidgets.QMainWindow): _config_file: Path | None _config: dict _text_osam_session: OsamSession | None = None _is_changed: bool = False _copied_shapes: list[Shape] _zoom_mode: _ZoomMode _prev_opened_dir: str | None _canvas_widgets: _CanvasWidgets _status_bar: _StatusBarWidgets _docks: _DockWidgets _actions: _Actions _menus: _Menus _scalers: dict[_ZoomMode, Callable[[], float]] _label_dialog: LabelDialog _ai_annotation: AiAssistedAnnotationWidget _ai_text: AiTextToAnnotationWidget _output_dir: str | None _filename: str | None _image: QtGui.QImage _label_file: LabelFile | None _image_path: str | None _recent_files: list[str] _max_recent: int _other_data: dict | None _zoom_values: dict[str, tuple[_ZoomMode, int]] _brightness_contrast_values: dict[str, tuple[int | None, int | None]] _scroll_values: dict[Qt.Orientation, dict[str, float]] _default_state: QtCore.QByteArray def __init__( self, config_file: Path | None = None, config_overrides: dict | None = None, filename: str | None = None, output_dir: str | None = None, ) -> None: super().__init__() self.setWindowTitle(__appname__) self._config_file, self._config = self._load_config( config_file=config_file, config_overrides=config_overrides ) # set default shape colors Shape.line_color = QtGui.QColor(*self._config["shape"]["line_color"]) Shape.fill_color = QtGui.QColor(*self._config["shape"]["fill_color"]) Shape.select_line_color = QtGui.QColor( *self._config["shape"]["select_line_color"] ) Shape.select_fill_color = QtGui.QColor( *self._config["shape"]["select_fill_color"] ) Shape.vertex_fill_color = QtGui.QColor( *self._config["shape"]["vertex_fill_color"] ) Shape.hvertex_fill_color = QtGui.QColor( *self._config["shape"]["hvertex_fill_color"] ) # Set point size from config file Shape.point_size = self._config["shape"]["point_size"] self._copied_shapes = [] self._label_dialog = LabelDialog( parent=self, labels=self._config["labels"], sort_labels=self._config["sort_labels"], show_text_field=self._config["show_label_text_field"], completion=self._config["label_completion"], fit_to_content=self._config["fit_to_content"], flags=self._config["label_flags"], ) self._prev_opened_dir = None self._docks = self._setup_dock_widgets() self.setAcceptDrops(True) self._canvas_widgets = self._setup_canvas() self._actions = self._setup_actions() self._scalers = { _ZoomMode.FIT_WINDOW: self.scaleFitWindow, _ZoomMode.FIT_WIDTH: self.scaleFitWidth, _ZoomMode.MANUAL_ZOOM: lambda: 1, } self._menus = self._setup_menus() self._ai_annotation = AiAssistedAnnotationWidget( default_model=self._config["ai"]["default"], on_model_changed=self._canvas_widgets.canvas.set_ai_model_name, parent=self, ) self._ai_annotation.setEnabled(False) self._ai_text = AiTextToAnnotationWidget( on_submit=self._submit_ai_prompt, parent=self ) self._ai_text.setEnabled(False) self._setup_toolbars() self._status_bar = self._setup_status_bar() self._setup_app_state(output_dir=output_dir, filename=filename) self.updateFileMenu() self._canvas_widgets.zoom_widget.valueChanged.connect(self._paint_canvas) self.populateModeActions() def _setup_actions(self) -> _Actions: action = functools.partial(utils.newAction, self) shortcuts = self._config["shortcuts"] about = action( text=f"&About {__appname__}", slot=functools.partial( QMessageBox.about, self, f"About {__appname__}", f"""
{__appname__}
Image Polygonal Annotation with Python
Version: {__version__}
Author: Kentaro Wada
Homepage | Documentation | Troubleshooting
""", ), ) save = action( text=self.tr("&Save\n"), slot=self.saveFile, shortcut=shortcuts["save"], icon="floppy-disk.svg", tip=self.tr("Save labels to file"), enabled=False, ) save_as = action( text=self.tr("&Save As"), slot=self.saveFileAs, shortcut=shortcuts["save_as"], icon="floppy-disk.svg", tip=self.tr("Save labels to a different file"), enabled=False, ) save_auto = action( text=self.tr("Save &Automatically"), tip=self.tr("Save automatically"), checkable=True, enabled=True, ) save_auto.setChecked(self._config["auto_save"]) save_with_image_data = action( text=self.tr("Save With Image Data"), slot=self.enableSaveImageWithData, tip=self.tr("Save image data in label file"), checkable=True, checked=self._config["with_image_data"], ) change_output_dir = action( text=self.tr("&Change Output Dir"), slot=self.changeOutputDirDialog, shortcut=shortcuts["save_to"], icon="folders.svg", tip=self.tr("Change where annotations are loaded/saved"), ) open_ = action( text=self.tr("&Open\n"), slot=self._open_file_with_dialog, shortcut=shortcuts["open"], icon="folder-open.svg", tip=self.tr("Open image or label file"), ) open_dir = action( text=self.tr("Open Dir"), slot=self._open_dir_with_dialog, shortcut=shortcuts["open_dir"], icon="folder-open.svg", tip=self.tr("Open Dir"), ) close = action( text=self.tr("&Close"), slot=self.closeFile, shortcut=shortcuts["close"], icon="x-circle.svg", tip=self.tr("Close current file"), ) delete_file = action( text=self.tr("&Delete File"), slot=self.deleteFile, shortcut=shortcuts["delete_file"], icon="file-x.svg", tip=self.tr("Delete current label file"), enabled=False, ) toggle_keep_prev_mode = action( text=self.tr("Keep Previous Annotation"), slot=self.toggleKeepPrevMode, shortcut=shortcuts["toggle_keep_prev_mode"], icon=None, tip=self.tr('Toggle "keep previous annotation" mode'), checkable=True, ) toggle_keep_prev_mode.setChecked(self._config["keep_prev"]) toggle_keep_prev_brightness_contrast = action( text=self.tr("Keep Previous Brightness/Contrast"), slot=lambda: self._config.__setitem__( "keep_prev_brightness_contrast", not self._config["keep_prev_brightness_contrast"], ), checkable=True, checked=self._config["keep_prev_brightness_contrast"], ) delete = action( self.tr("Delete Shapes"), self.deleteSelectedShape, shortcuts["delete_shape"], icon="trash.svg", tip=self.tr("Delete the selected shapes"), enabled=False, ) edit = action( self.tr("&Edit Label"), self._edit_label, shortcuts["edit_label"], icon="note-pencil.svg", tip=self.tr("Modify the label of the selected shape"), enabled=False, ) duplicate = action( self.tr("Duplicate Shapes"), self.duplicateSelectedShape, shortcuts["duplicate_shape"], icon="copy.svg", tip=self.tr("Create a duplicate of the selected shapes"), enabled=False, ) copy = action( self.tr("Copy Shapes"), self.copySelectedShape, shortcuts["copy_shape"], "copy_clipboard", self.tr("Copy selected shapes to clipboard"), enabled=False, ) paste = action( self.tr("Paste Shapes"), self.pasteSelectedShape, shortcuts["paste_shape"], "paste", self.tr("Paste copied shapes"), enabled=False, ) undo_last_point = action( self.tr("Undo last point"), self._canvas_widgets.canvas.undoLastPoint, shortcuts["undo_last_point"], icon="arrow-u-up-left.svg", tip=self.tr("Undo last drawn point"), enabled=False, ) undo = action( self.tr("Undo\n"), self.undoShapeEdit, shortcuts["undo"], icon="arrow-u-up-left.svg", tip=self.tr("Undo last add and edit of shape"), enabled=False, ) remove_point = action( text=self.tr("Remove Selected Point"), slot=self.removeSelectedPoint, shortcut=shortcuts["remove_selected_point"], icon="trash.svg", tip=self.tr("Remove selected point from polygon"), enabled=False, ) create_mode = action( text=self.tr("Create Polygons"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="polygon"), shortcut=shortcuts["create_polygon"], icon="polygon.svg", tip=self.tr("Start drawing polygons"), enabled=False, ) edit_mode = action( self.tr("Edit Shapes"), lambda: self._switch_canvas_mode(edit=True), shortcuts["edit_shape"], icon="note-pencil.svg", tip=self.tr("Move and edit the selected shapes"), enabled=False, ) create_rectangle_mode = action( text=self.tr("Create Rectangle"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="rectangle"), shortcut=shortcuts["create_rectangle"], icon="rectangle.svg", tip=self.tr("Start drawing rectangles"), enabled=False, ) create_circle_mode = action( text=self.tr("Create Circle"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="circle"), shortcut=shortcuts["create_circle"], icon="circle.svg", tip=self.tr("Start drawing circles"), enabled=False, ) create_line_mode = action( text=self.tr("Create Line"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="line"), shortcut=shortcuts["create_line"], icon="line-segment.svg", tip=self.tr("Start drawing lines"), enabled=False, ) create_point_mode = action( text=self.tr("Create Point"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="point"), shortcut=shortcuts["create_point"], icon="circles-four.svg", tip=self.tr("Start drawing points"), enabled=False, ) create_line_strip_mode = action( text=self.tr("Create LineStrip"), slot=lambda: self._switch_canvas_mode(edit=False, createMode="linestrip"), shortcut=shortcuts["create_linestrip"], icon="line-segments.svg", tip=self.tr("Start drawing linestrip. Ctrl+LeftClick ends creation."), enabled=False, ) create_ai_polygon_mode = action( self.tr("Create AI-Polygon"), lambda: self._switch_canvas_mode(edit=False, createMode="ai_polygon"), None, "ai-polygon.svg", self.tr("Start drawing ai_polygon. Ctrl+LeftClick ends creation."), enabled=False, ) create_ai_mask_mode = action( self.tr("Create AI-Mask"), lambda: self._switch_canvas_mode(edit=False, createMode="ai_mask"), None, "ai-mask.svg", self.tr("Start drawing ai_mask. Ctrl+LeftClick ends creation."), enabled=False, ) open_next_img = action( text=self.tr("&Next Image"), slot=self._open_next_image, shortcut=shortcuts["open_next"], icon="arrow-fat-right.svg", tip=self.tr("Open next (hold Ctl+Shift to copy labels)"), enabled=False, ) open_prev_img = action( text=self.tr("&Prev Image"), slot=self._open_prev_image, shortcut=shortcuts["open_prev"], icon="arrow-fat-left.svg", tip=self.tr("Open prev (hold Ctl+Shift to copy labels)"), enabled=False, ) keep_prev_scale = action( self.tr("&Keep Previous Scale"), self.enableKeepPrevScale, tip=self.tr("Keep previous zoom scale"), checkable=True, checked=self._config["keep_prev_scale"], enabled=True, ) fit_window = action( self.tr("&Fit Window"), self.setFitWindow, shortcuts["fit_window"], icon="frame-corners.svg", tip=self.tr("Zoom follows window size"), checkable=True, enabled=False, ) fit_width = action( self.tr("Fit &Width"), self.setFitWidth, shortcuts["fit_width"], icon="frame-arrows-horizontal.svg", tip=self.tr("Zoom follows window width"), checkable=True, enabled=False, ) brightness_contrast = action( self.tr("&Brightness Contrast"), self.brightnessContrast, None, "brightness-contrast.svg", self.tr("Adjust brightness and contrast"), enabled=False, ) zoom_in = action( self.tr("Zoom &In"), lambda _: self._add_zoom(increment=1.1), shortcuts["zoom_in"], icon="magnifying-glass-minus.svg", tip=self.tr("Increase zoom level"), enabled=False, ) zoom_out = action( self.tr("&Zoom Out"), lambda _: self._add_zoom(increment=0.9), shortcuts["zoom_out"], icon="magnifying-glass-plus.svg", tip=self.tr("Decrease zoom level"), enabled=False, ) zoom_org = action( self.tr("&Original size"), self._set_zoom_to_original, shortcuts["zoom_to_original"], icon="image-square.svg", tip=self.tr("Zoom to original size"), enabled=False, ) reset_layout = action( text=self.tr("Reset Layout"), slot=self._reset_layout, icon="layout-duotone.svg", ) fill_drawing = action( self.tr("Fill Drawing Polygon"), self._canvas_widgets.canvas.setFillDrawing, None, icon="paint-bucket.svg", tip=self.tr("Fill polygon while drawing"), checkable=True, enabled=True, ) if self._config["canvas"]["fill_drawing"]: fill_drawing.trigger() hide_all = action( self.tr("&Hide\nShapes"), functools.partial(self.toggleShapes, False), shortcuts["hide_all_shapes"], icon="eye.svg", tip=self.tr("Hide all shapes"), enabled=False, ) show_all = action( self.tr("&Show\nShapes"), functools.partial(self.toggleShapes, True), shortcuts["show_all_shapes"], icon="eye.svg", tip=self.tr("Show all shapes"), enabled=False, ) toggle_all = action( self.tr("&Toggle\nShapes"), functools.partial(self.toggleShapes, None), shortcuts["toggle_all_shapes"], icon="eye.svg", tip=self.tr("Toggle all shapes"), enabled=False, ) zoom_widget_action = QtWidgets.QWidgetAction(self) zoom_box_layout = QtWidgets.QVBoxLayout() zoom_label = QtWidgets.QLabel(self.tr("Zoom")) zoom_label.setAlignment(Qt.AlignCenter) zoom_box_layout.addWidget(zoom_label) zoom_box_layout.addWidget(self._canvas_widgets.zoom_widget) zoom_widget_action.setDefaultWidget(QtWidgets.QWidget()) zoom_widget_action.defaultWidget().setLayout(zoom_box_layout) self._canvas_widgets.zoom_widget.setWhatsThis( str( self.tr( "Zoom in or out of the image. Also accessible with " "{} and {} from the canvas." ) ).format( utils.fmtShortcut(f"{shortcuts['zoom_in']},{shortcuts['zoom_out']}"), utils.fmtShortcut(self.tr("Ctrl+Wheel")), ) ) self._canvas_widgets.zoom_widget.setEnabled(False) self._zoom_mode = _ZoomMode.FIT_WINDOW fit_window.setChecked(Qt.Checked) self._canvas_widgets.canvas.vertexSelected.connect(remove_point.setEnabled) draw = [ ("polygon", create_mode), ("rectangle", create_rectangle_mode), ("circle", create_circle_mode), ("point", create_point_mode), ("line", create_line_mode), ("linestrip", create_line_strip_mode), ("ai_polygon", create_ai_polygon_mode), ("ai_mask", create_ai_mask_mode), ] zoom = ( self._canvas_widgets.zoom_widget, zoom_in, zoom_out, zoom_org, fit_window, fit_width, ) on_load_active = ( close, create_mode, create_rectangle_mode, create_circle_mode, create_line_mode, create_point_mode, create_line_strip_mode, create_ai_polygon_mode, create_ai_mask_mode, brightness_contrast, ) on_shapes_present = (save_as, hide_all, show_all, toggle_all) context_menu = ( *[draw_action for _, draw_action in draw], edit_mode, edit, duplicate, copy, paste, delete, undo, undo_last_point, remove_point, ) edit_menu = ( edit, duplicate, copy, paste, delete, None, undo, undo_last_point, None, remove_point, None, toggle_keep_prev_mode, ) return _Actions( about=about, save=save, save_as=save_as, save_auto=save_auto, save_with_image_data=save_with_image_data, change_output_dir=change_output_dir, open=open_, close=close, delete_file=delete_file, toggle_keep_prev_mode=toggle_keep_prev_mode, toggle_keep_prev_brightness_contrast=toggle_keep_prev_brightness_contrast, delete=delete, edit=edit, duplicate=duplicate, copy=copy, paste=paste, undo_last_point=undo_last_point, undo=undo, remove_point=remove_point, create_mode=create_mode, edit_mode=edit_mode, create_rectangle_mode=create_rectangle_mode, create_circle_mode=create_circle_mode, create_line_mode=create_line_mode, create_point_mode=create_point_mode, create_line_strip_mode=create_line_strip_mode, create_ai_polygon_mode=create_ai_polygon_mode, create_ai_mask_mode=create_ai_mask_mode, open_next_img=open_next_img, open_prev_img=open_prev_img, keep_prev_scale=keep_prev_scale, fit_window=fit_window, fit_width=fit_width, brightness_contrast=brightness_contrast, zoom_in=zoom_in, zoom_out=zoom_out, zoom_org=zoom_org, reset_layout=reset_layout, fill_drawing=fill_drawing, hide_all=hide_all, show_all=show_all, toggle_all=toggle_all, open_dir=open_dir, zoom_widget_action=zoom_widget_action, draw=draw, zoom=zoom, on_load_active=on_load_active, on_shapes_present=on_shapes_present, context_menu=context_menu, edit_menu=edit_menu, ) def _setup_menus(self) -> _Menus: action = functools.partial(utils.newAction, self) shortcuts = self._config["shortcuts"] quit_ = action( text=self.tr("&Quit"), slot=self.close, shortcut=shortcuts["quit"], icon=None, tip=self.tr("Quit application"), ) open_config = action( text=self.tr("Preferences…"), slot=self._open_config_file, shortcut="Ctrl+," if platform.system() == "Darwin" else "Ctrl+Shift+,", icon=None, tip=self.tr("Open config file in text editor"), ) open_config.setMenuRole(QtWidgets.QAction.PreferencesRole) help_ = action( self.tr("&Tutorial"), self.tutorial, icon="question.svg", tip=self.tr("Show tutorial page"), ) file_menu = self.menu(self.tr("&File")) edit_menu = self.menu(self.tr("&Edit")) view_menu = self.menu(self.tr("&View")) help_menu = self.menu(self.tr("&Help")) recent_files = QtWidgets.QMenu(self.tr("Open &Recent")) label_menu = QtWidgets.QMenu() utils.addActions(label_menu, (self._actions.edit, self._actions.delete)) self._docks.label_list.setContextMenuPolicy(Qt.CustomContextMenu) self._docks.label_list.customContextMenuRequested.connect(self.popLabelListMenu) utils.addActions( file_menu, ( self._actions.open, self._actions.open_next_img, self._actions.open_prev_img, self._actions.open_dir, recent_files, self._actions.save, self._actions.save_as, self._actions.save_auto, self._actions.change_output_dir, self._actions.save_with_image_data, self._actions.close, self._actions.delete_file, None, open_config, None, quit_, ), ) utils.addActions(help_menu, (help_, self._actions.about)) utils.addActions( view_menu, ( self._docks.flag_dock.toggleViewAction(), self._docks.label_dock.toggleViewAction(), self._docks.shape_dock.toggleViewAction(), self._docks.file_dock.toggleViewAction(), None, self._actions.reset_layout, None, self._actions.fill_drawing, None, self._actions.hide_all, self._actions.show_all, self._actions.toggle_all, None, self._actions.zoom_in, self._actions.zoom_out, self._actions.zoom_org, self._actions.keep_prev_scale, None, self._actions.fit_window, self._actions.fit_width, None, self._actions.brightness_contrast, self._actions.toggle_keep_prev_brightness_contrast, ), ) file_menu.aboutToShow.connect(self.updateFileMenu) utils.addActions( self._canvas_widgets.canvas.menus[0], self._actions.context_menu ) utils.addActions( self._canvas_widgets.canvas.menus[1], ( action("&Copy here", self.copyShape), action("&Move here", self.moveShape), ), ) return _Menus( file=file_menu, edit=edit_menu, view=view_menu, help=help_menu, recent_files=recent_files, label_list=label_menu, ) def _setup_toolbars(self) -> None: select_ai_model = QtWidgets.QWidgetAction(self) select_ai_model.setDefaultWidget(self._ai_annotation) ai_prompt_action = QtWidgets.QWidgetAction(self) ai_prompt_action.setDefaultWidget(self._ai_text) self.addToolBar( Qt.TopToolBarArea, ToolBar( title="Tools", actions=[ self._actions.open, self._actions.open_dir, self._actions.open_prev_img, self._actions.open_next_img, self._actions.save, self._actions.delete_file, None, self._actions.edit_mode, self._actions.duplicate, self._actions.delete, self._actions.undo, self._actions.brightness_contrast, None, self._actions.fit_window, self._actions.zoom_widget_action, None, select_ai_model, None, ai_prompt_action, ], font_base=self.font(), ), ) self.addToolBar( Qt.LeftToolBarArea, ToolBar( title="CreateShapeTools", actions=[a for _, a in self._actions.draw], orientation=Qt.Vertical, button_style=Qt.ToolButtonTextUnderIcon, font_base=self.font(), ), ) def _setup_app_state( self, *, output_dir: str | None, filename: str | None, ) -> None: self._output_dir = output_dir self._image = QtGui.QImage() self._label_file = None self._image_path = None self._max_recent = 7 self._other_data = None self._zoom_values = {} self._brightness_contrast_values = {} self._scroll_values = { Qt.Horizontal: {}, Qt.Vertical: {}, } if self._config["file_search"]: self._docks.file_search.setText(self._config["file_search"]) self._default_state = self.saveState() # # XXX: Could be completely declarative. # Restore application settings. self.settings = QtCore.QSettings("labelme", "labelme") # # Bump this when dock/toolbar layout changes to reset window state # for users upgrading from an older version. SETTINGS_VERSION: int = 1 if self.settings.value("settingsVersion", 0, type=int) != SETTINGS_VERSION: self._reset_layout() self.settings.setValue("settingsVersion", SETTINGS_VERSION) # self._recent_files = self.settings.value("recentFiles", []) or [] self.resize(self.settings.value("window/size", QtCore.QSize(900, 500))) self.move(self.settings.value("window/position", QtCore.QPoint(0, 0))) self.restoreState(self.settings.value("window/state", QtCore.QByteArray())) # Recover window position when the saved screen is no longer connected. if not any( s.availableGeometry().intersects(self.frameGeometry()) for s in QtWidgets.QApplication.screens() ) and (primary_screen := QtWidgets.QApplication.primaryScreen()): self.move(primary_screen.availableGeometry().topLeft()) if filename: if osp.isdir(filename): self._import_images_from_dir( root_dir=filename, pattern=self._docks.file_search.text() ) self._open_next_image() else: self._load_file(filename=filename) else: self._filename = None def _setup_status_bar(self) -> _StatusBarWidgets: message = QtWidgets.QLabel(self.tr("%s started.") % __appname__) stats = StatusStats() self.statusBar().addWidget(message, 1) self.statusBar().addWidget(stats, 0) self.statusBar().show() return _StatusBarWidgets(message=message, stats=stats) def _setup_canvas(self) -> _CanvasWidgets: zoom_widget = ZoomWidget() canvas = Canvas( epsilon=self._config["epsilon"], double_click=self._config["canvas"]["double_click"], num_backups=self._config["canvas"]["num_backups"], crosshair=self._config["canvas"]["crosshair"], ) canvas.zoomRequest.connect(self._zoom_requested) canvas.mouseMoved.connect(self._update_status_stats) canvas.statusUpdated.connect( lambda text: self._status_bar.message.setText(text) ) scroll_area = QtWidgets.QScrollArea() scroll_area.setWidget(canvas) scroll_area.setWidgetResizable(True) scroll_bars = { Qt.Vertical: scroll_area.verticalScrollBar(), Qt.Horizontal: scroll_area.horizontalScrollBar(), } canvas.scrollRequest.connect(self.scrollRequest) canvas.newShape.connect(self.newShape) canvas.shapeMoved.connect(self.setDirty) canvas.selectionChanged.connect(self.shapeSelectionChanged) canvas.drawingPolygon.connect(self.toggleDrawingSensitive) self.setCentralWidget(scroll_area) return _CanvasWidgets( canvas=canvas, zoom_widget=zoom_widget, scroll_bars=scroll_bars, ) def _setup_dock_widgets(self) -> _DockWidgets: flag_list = QtWidgets.QListWidget() flag = QtWidgets.QDockWidget(self.tr("Flags"), self) flag.setObjectName("Flags") if self._config["flags"]: self._load_flags( flags={k: False for k in self._config["flags"]}, widget=flag_list, ) flag.setWidget(flag_list) flag_list.itemChanged.connect(self.setDirty) label_list = LabelListWidget() label_list.itemSelectionChanged.connect(self._label_selection_changed) label_list.itemDoubleClicked.connect(self._edit_label) label_list.itemChanged.connect(self.labelItemChanged) label_list.itemDropped.connect(self.labelOrderChanged) shape = QtWidgets.QDockWidget(self.tr("Annotation List"), self) shape.setObjectName("Labels") shape.setWidget(label_list) unique_label_list = UniqueLabelQListWidget() unique_label_list.setToolTip( self.tr("Select label to start annotating for it. Press 'Esc' to deselect.") ) if self._config["labels"]: for lbl in self._config["labels"]: unique_label_list.add_label_item( label=lbl, color=self._get_rgb_by_label( label=lbl, unique_label_list=unique_label_list ), ) label = QtWidgets.QDockWidget(self.tr("Label List"), self) label.setObjectName("Label List") label.setWidget(unique_label_list) file_search = QtWidgets.QLineEdit() file_search.setPlaceholderText(self.tr("Search Filename")) file_search.textChanged.connect(self.fileSearchChanged) file_list = QtWidgets.QListWidget() file_list.itemSelectionChanged.connect(self.fileSelectionChanged) file_list_layout = QtWidgets.QVBoxLayout() file_list_layout.setContentsMargins(0, 0, 0, 0) file_list_layout.setSpacing(0) file_list_layout.addWidget(file_search) file_list_layout.addWidget(file_list) file = QtWidgets.QDockWidget(self.tr("File List"), self) file.setObjectName("Files") file_list_container = QtWidgets.QWidget() file_list_container.setLayout(file_list_layout) file.setWidget(file_list_container) for config_key, dock_widget in [ ("flag_dock", flag), ("label_dock", label), ("shape_dock", shape), ("file_dock", file), ]: features = QtWidgets.QDockWidget.DockWidgetFeatures() if self._config[config_key]["closable"]: features = features | QtWidgets.QDockWidget.DockWidgetClosable if self._config[config_key]["floatable"]: features = features | QtWidgets.QDockWidget.DockWidgetFloatable if self._config[config_key]["movable"]: features = features | QtWidgets.QDockWidget.DockWidgetMovable dock_widget.setFeatures(features) if self._config[config_key]["show"] is False: dock_widget.setVisible(False) self.addDockWidget(Qt.RightDockWidgetArea, dock_widget) return _DockWidgets( flag_dock=flag, flag_list=flag_list, shape_dock=shape, label_list=label_list, label_dock=label, unique_label_list=unique_label_list, file_dock=file, file_search=file_search, file_list=file_list, ) def _load_config( self, config_file: Path | None, config_overrides: dict | None ) -> tuple[Path | None, dict]: try: config = load_config( config_file=config_file, config_overrides=config_overrides or {} ) except ValueError as e: msg_box = QMessageBox(self) msg_box.setIcon(QMessageBox.Warning) msg_box.setWindowTitle(self.tr("Configuration Errors")) msg_box.setText( self.tr( "Errors were found while loading the configuration. " "Please review the errors below and reload your configuration or " "ignore the erroneous lines." ) ) msg_box.setInformativeText(str(e)) msg_box.setStandardButtons(QMessageBox.Ignore) msg_box.setModal(False) msg_box.show() config_file = None config_overrides = {} config = load_config( config_file=config_file, config_overrides=config_overrides ) return config_file, config def menu(self, title, actions=None): menu = self.menuBar().addMenu(title) if actions: utils.addActions(menu, actions) return menu # Support Functions def noShapes(self) -> bool: return not len(self._docks.label_list) def populateModeActions(self) -> None: self._canvas_widgets.canvas.menus[0].clear() utils.addActions( self._canvas_widgets.canvas.menus[0], self._actions.context_menu ) self._menus.edit.clear() actions = ( *[draw_action for _, draw_action in self._actions.draw], self._actions.edit_mode, *self._actions.edit_menu, ) utils.addActions(self._menus.edit, actions) def _get_window_title(self, dirty: bool) -> str: window_title: str = __appname__ if self._image_path: window_title = f"{window_title} - {self._image_path}" if self._docks.file_list.count() and self._docks.file_list.currentItem(): window_title = ( f"{window_title} " f"[{self._docks.file_list.currentRow() + 1}" f"/{self._docks.file_list.count()}]" ) if dirty: window_title = f"{window_title}*" return window_title def setDirty(self) -> None: # Even if we autosave the file, we keep the ability to undo self._actions.undo.setEnabled(self._canvas_widgets.canvas.isShapeRestorable) if self._config["auto_save"] or self._actions.save_auto.isChecked(): assert self._image_path label_file = f"{osp.splitext(self._image_path)[0]}.json" if self._output_dir: label_file_without_path = osp.basename(label_file) label_file = osp.join(self._output_dir, label_file_without_path) self.saveLabels(label_file) return self._is_changed = True self._actions.save.setEnabled(True) self.setWindowTitle(self._get_window_title(dirty=True)) def setClean(self) -> None: self._is_changed = False self._actions.save.setEnabled(False) for _, action in self._actions.draw: action.setEnabled(True) self.setWindowTitle(self._get_window_title(dirty=False)) if self.hasLabelFile(): self._actions.delete_file.setEnabled(True) else: self._actions.delete_file.setEnabled(False) def toggleActions(self, value: bool = True) -> None: """Enable/Disable widgets which depend on an opened image.""" for z in self._actions.zoom: z.setEnabled(value) for action in self._actions.on_load_active: action.setEnabled(value) def queueEvent(self, function: Callable[[], None]) -> None: QtCore.QTimer.singleShot(0, function) def show_status_message(self, message: str, delay: int = 500) -> None: self.statusBar().showMessage(message, delay) def _submit_ai_prompt(self, _) -> None: if ( self._canvas_widgets.canvas.createMode not in _AI_TEXT_TO_ANNOTATION_CREATE_MODE_TO_SHAPE_TYPE ): logger.warning( "Unsupported createMode={!r}", self._canvas_widgets.canvas.createMode ) return shape_type: Literal["rectangle", "polygon", "mask"] = ( _AI_TEXT_TO_ANNOTATION_CREATE_MODE_TO_SHAPE_TYPE[ self._canvas_widgets.canvas.createMode ] ) texts = self._ai_text.get_text_prompt().split(",") model_name: str = self._ai_text.get_model_name() model_type = osam.apis.get_model_type_by_name(model_name) if not (_is_already_downloaded := model_type.get_size() is not None): if not download_ai_model(model_name=model_name, parent=self): return if ( self._text_osam_session is None or self._text_osam_session.model_name != model_name ): self._text_osam_session = OsamSession(model_name=model_name) boxes, scores, labels, masks = bbox_from_text.get_bboxes_from_texts( session=self._text_osam_session, image=utils.img_qt_to_arr(self._image)[:, :, :3], image_id=str(hash(self._image_path)), texts=texts, ) SCORE_FOR_EXISTING_SHAPE: float = 1.01 for shape in self._canvas_widgets.canvas.shapes: if shape.shape_type != shape_type or shape.label not in texts: continue points: NDArray[np.float64] = np.array( [[p.x(), p.y()] for p in shape.points] ) xmin, ymin = points.min(axis=0) xmax, ymax = points.max(axis=0) box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32) boxes = np.r_[boxes, [box]] scores = np.r_[scores, [SCORE_FOR_EXISTING_SHAPE]] labels = np.r_[labels, [texts.index(shape.label)]] boxes, scores, labels, indices = bbox_from_text.nms_bboxes( boxes=boxes, scores=scores, labels=labels, iou_threshold=self._ai_text.get_iou_threshold(), score_threshold=self._ai_text.get_score_threshold(), max_num_detections=100, ) is_new = scores != SCORE_FOR_EXISTING_SHAPE boxes = boxes[is_new] scores = scores[is_new] labels = labels[is_new] indices = indices[is_new] if masks is not None: masks = masks[indices] del indices shapes: list[Shape] = bbox_from_text.get_shapes_from_bboxes( boxes=boxes, scores=scores, labels=labels, texts=texts, masks=masks, shape_type=shape_type, ) self._canvas_widgets.canvas.storeShapes() self._load_shapes(shapes, replace=False) self.setDirty() def resetState(self) -> None: self._docks.label_list.clear() self._filename = None self._image_path = None self.imageData = None self._label_file = None self._other_data = None self._canvas_widgets.canvas.resetState() def currentItem(self) -> LabelListWidgetItem | None: items = self._docks.label_list.selectedItems() if items: return items[0] return None def addRecentFile(self, filename: str) -> None: if filename in self._recent_files: self._recent_files.remove(filename) elif len(self._recent_files) >= self._max_recent: self._recent_files.pop() self._recent_files.insert(0, filename) # Callbacks def undoShapeEdit(self) -> None: self._canvas_widgets.canvas.restoreShape() self._docks.label_list.clear() self._load_shapes(self._canvas_widgets.canvas.shapes) self._actions.undo.setEnabled(self._canvas_widgets.canvas.isShapeRestorable) def tutorial(self): url = "https://github.com/labelmeai/labelme/tree/main/examples/tutorial" # NOQA webbrowser.open(url) def toggleDrawingSensitive(self, drawing=True): """Toggle drawing sensitive. In the middle of drawing, toggling between modes should be disabled. """ self._actions.edit_mode.setEnabled(not drawing) self._actions.undo_last_point.setEnabled(drawing) self._actions.undo.setEnabled(not drawing) self._actions.delete.setEnabled(not drawing) def _switch_canvas_mode( self, edit: bool = True, createMode: str | None = None ) -> None: self._canvas_widgets.canvas.setEditing(edit) if createMode is not None: self._canvas_widgets.canvas.createMode = createMode if edit: for _, draw_action in self._actions.draw: draw_action.setEnabled(True) else: for draw_mode, draw_action in self._actions.draw: draw_action.setEnabled(createMode != draw_mode) self._actions.edit_mode.setEnabled(not edit) self._ai_text.setEnabled( not edit and createMode in _AI_TEXT_TO_ANNOTATION_CREATE_MODE_TO_SHAPE_TYPE ) self._ai_annotation.setEnabled( not edit and createMode in ("ai_polygon", "ai_mask") ) def updateFileMenu(self): current = self._filename def exists(filename): return osp.exists(str(filename)) menu = self._menus.recent_files menu.clear() files = [f for f in self._recent_files if f != current and exists(f)] for i, f in enumerate(files): icon = utils.newIcon("labels") action = QtWidgets.QAction( icon, f"&{i + 1} {QtCore.QFileInfo(f).fileName()}", self ) action.triggered.connect(functools.partial(self.loadRecent, f)) menu.addAction(action) def popLabelListMenu(self, point: QtCore.QPoint) -> None: self._menus.label_list.exec(self._docks.label_list.mapToGlobal(point)) # type: ignore[invalid-argument-type] def validateLabel(self, label): # no validation if self._config["validate_label"] is None: return True for i in range(self._docks.unique_label_list.count()): label_i = self._docks.unique_label_list.item(i).data(Qt.UserRole) # type: ignore[attr-defined,union-attr] if self._config["validate_label"] in ["exact"]: if label_i == label: return True return False def _edit_label(self, value=None): items = self._docks.label_list.selectedItems() if not items: logger.warning("No label is selected, so cannot edit label.") return shape = items[0].shape() if len(items) == 1: edit_text = True edit_flags = True edit_group_id = True edit_description = True else: edit_text = all(item.shape().label == shape.label for item in items[1:]) edit_flags = all(item.shape().flags == shape.flags for item in items[1:]) edit_group_id = all( item.shape().group_id == shape.group_id for item in items[1:] ) edit_description = all( item.shape().description == shape.description for item in items[1:] ) if not edit_text: self._label_dialog.edit.setDisabled(True) self._label_dialog.labelList.setDisabled(True) if not edit_group_id: self._label_dialog.edit_group_id.setDisabled(True) if not edit_description: self._label_dialog.editDescription.setDisabled(True) text, flags, group_id, description = self._label_dialog.popUp( text=shape.label if edit_text else "", flags=shape.flags if edit_flags else None, group_id=shape.group_id if edit_group_id else None, description=shape.description if edit_description else None, flags_disabled=not edit_flags, ) if not edit_text: self._label_dialog.edit.setDisabled(False) self._label_dialog.labelList.setDisabled(False) if not edit_group_id: self._label_dialog.edit_group_id.setDisabled(False) if not edit_description: self._label_dialog.editDescription.setDisabled(False) if text is None: assert flags is None assert group_id is None assert description is None return if not self.validateLabel(text): self.errorMessage( self.tr("Invalid label"), self.tr("Invalid label '{}' with validation type '{}'").format( text, self._config["validate_label"] ), ) return self._canvas_widgets.canvas.storeShapes() for item in items: shape: Shape = item.shape() # type: ignore[no-redef] if edit_text: shape.label = text if edit_flags: shape.flags = flags if edit_group_id: shape.group_id = group_id if edit_description: shape.description = description self._update_shape_color(shape) if shape.group_id is None: r, g, b = shape.fill_color.getRgb()[:3] item.setText( f"{html.escape(shape.label)} " f'●' ) else: item.setText(f"{shape.label} ({shape.group_id})") self.setDirty() if self._docks.unique_label_list.find_label_item(shape.label) is None: self._docks.unique_label_list.add_label_item( label=shape.label, color=self._get_rgb_by_label( label=shape.label, unique_label_list=self._docks.unique_label_list, ), ) def fileSearchChanged(self): self._import_images_from_dir( root_dir=self._prev_opened_dir, pattern=self._docks.file_search.text() ) def fileSelectionChanged(self) -> None: items = self._docks.file_list.selectedItems() if not items: return item = items[0] if not self._can_continue(): return curr_index = self.imageList.index(str(item.text())) if curr_index < len(self.imageList): filename = self.imageList[curr_index] if filename: self._load_file(filename) # React to canvas signals. def shapeSelectionChanged(self, selected_shapes: list[Shape]) -> None: self._docks.label_list.itemSelectionChanged.disconnect( self._label_selection_changed ) for shape in self._canvas_widgets.canvas.selectedShapes: shape.selected = False self._docks.label_list.clearSelection() self._canvas_widgets.canvas.selectedShapes = selected_shapes for shape in self._canvas_widgets.canvas.selectedShapes: shape.selected = True item = self._docks.label_list.findItemByShape(shape) self._docks.label_list.selectItem(item) self._docks.label_list.scrollToItem(item) self._docks.label_list.itemSelectionChanged.connect( self._label_selection_changed ) n_selected = len(selected_shapes) > 0 self._actions.delete.setEnabled(n_selected) self._actions.duplicate.setEnabled(n_selected) self._actions.copy.setEnabled(n_selected) self._actions.edit.setEnabled(n_selected) def addLabel(self, shape: Shape) -> None: assert shape.label is not None if shape.group_id is None: text = shape.label else: text = f"{shape.label} ({shape.group_id})" label_list_item = LabelListWidgetItem(text, shape) self._docks.label_list.addItem(label_list_item) if self._docks.unique_label_list.find_label_item(shape.label) is None: self._docks.unique_label_list.add_label_item( label=shape.label, color=self._get_rgb_by_label( label=shape.label, unique_label_list=self._docks.unique_label_list, ), ) self._label_dialog.addLabelHistory(shape.label) for action in self._actions.on_shapes_present: action.setEnabled(True) self._update_shape_color(shape) r, g, b = shape.fill_color.getRgb()[:3] label_list_item.setText( f'{html.escape(text)} ●' ) def _update_shape_color(self, shape: Shape) -> None: assert shape.label is not None r, g, b = self._get_rgb_by_label( shape.label, unique_label_list=self._docks.unique_label_list ) shape.line_color = QtGui.QColor(r, g, b) shape.vertex_fill_color = QtGui.QColor(r, g, b) shape.hvertex_fill_color = QtGui.QColor(255, 255, 255) shape.fill_color = QtGui.QColor(r, g, b, 128) shape.select_line_color = QtGui.QColor(255, 255, 255) shape.select_fill_color = QtGui.QColor(r, g, b, 155) def _get_rgb_by_label( self, label: str, unique_label_list: UniqueLabelQListWidget, ) -> tuple[int, int, int]: if self._config["shape_color"] == "auto": item = unique_label_list.find_label_item(label) item_index: int = ( unique_label_list.indexFromItem(item).row() if item else unique_label_list.count() ) label_id: int = ( 1 # skip black color by default + item_index + self._config["shift_auto_shape_color"] ) rgb: tuple[int, int, int] = tuple( LABEL_COLORMAP[label_id % len(LABEL_COLORMAP)].tolist() ) return rgb elif ( self._config["shape_color"] == "manual" and self._config["label_colors"] and label in self._config["label_colors"] ): if not ( len(self._config["label_colors"][label]) == 3 and all(0 <= c <= 255 for c in self._config["label_colors"][label]) ): raise ValueError( "Color for label must be 0-255 RGB tuple, but got: " f"{self._config['label_colors'][label]}" ) return tuple(self._config["label_colors"][label]) elif self._config["default_shape_color"]: return self._config["default_shape_color"] return (0, 255, 0) def remLabels(self, shapes: list[Shape]) -> None: for shape in shapes: item = self._docks.label_list.findItemByShape(shape) self._docks.label_list.removeItem(item) def _load_shapes(self, shapes: list[Shape], replace: bool = True) -> None: self._docks.label_list.itemSelectionChanged.disconnect( self._label_selection_changed ) shape: Shape for shape in shapes: self.addLabel(shape) self._docks.label_list.clearSelection() self._docks.label_list.itemSelectionChanged.connect( self._label_selection_changed ) self._canvas_widgets.canvas.loadShapes(shapes=shapes, replace=replace) def _load_shape_dicts(self, shape_dicts: list[ShapeDict]) -> None: shapes: list[Shape] = [] shape_dict: ShapeDict for shape_dict in shape_dicts: shape: Shape = Shape( label=shape_dict["label"], shape_type=shape_dict["shape_type"], group_id=shape_dict["group_id"], description=shape_dict["description"], mask=shape_dict["mask"], ) for x, y in shape_dict["points"]: shape.addPoint(QtCore.QPointF(x, y)) shape.close() default_flags = {} if self._config["label_flags"]: for pattern, keys in self._config["label_flags"].items(): if not isinstance(shape.label, str): logger.warning("shape.label is not str: {}", shape.label) continue if re.match(pattern, shape.label): for key in keys: default_flags[key] = False shape.flags = default_flags shape.flags.update(shape_dict["flags"]) shape.other_data = shape_dict["other_data"] shapes.append(shape) self._load_shapes(shapes=shapes) def _load_flags( self, flags: dict[str, bool], widget: QtWidgets.QListWidget, ) -> None: widget.clear() key: str flag: bool for key, flag in flags.items(): item: QtWidgets.QListWidgetItem = QtWidgets.QListWidgetItem(key) item.setFlags(item.flags() | Qt.ItemIsUserCheckable) item.setCheckState(Qt.Checked if flag else Qt.Unchecked) widget.addItem(item) def saveLabels(self, filename): lf = LabelFile() def format_shape(s): data = s.other_data.copy() data.update( dict( label=s.label, points=[(p.x(), p.y()) for p in s.points], group_id=s.group_id, description=s.description, shape_type=s.shape_type, flags=s.flags, mask=None if s.mask is None else utils.img_arr_to_b64(s.mask.astype(np.uint8)), ) ) return data shapes = [format_shape(item.shape()) for item in self._docks.label_list] flags = {} for i in range(self._docks.flag_list.count()): item = self._docks.flag_list.item(i) assert item key = item.text() flag = item.checkState() == Qt.Checked flags[key] = flag try: assert self._image_path imagePath = osp.relpath(self._image_path, osp.dirname(filename)) imageData = self.imageData if self._config["with_image_data"] else None if osp.dirname(filename) and not osp.exists(osp.dirname(filename)): os.makedirs(osp.dirname(filename)) lf.save( filename=filename, shapes=shapes, imagePath=imagePath, imageData=imageData, imageHeight=self._image.height(), imageWidth=self._image.width(), otherData=self._other_data, flags=flags, ) self._label_file = lf items = self._docks.file_list.findItems(self._image_path, Qt.MatchExactly) if len(items) > 0: if len(items) != 1: raise RuntimeError("There are duplicate files.") items[0].setCheckState(Qt.Checked) # disable allows next and previous image to proceed # self._filename = filename return True except LabelFileError as e: self.errorMessage( self.tr("Error saving label data"), self.tr("%s") % e ) return False def duplicateSelectedShape(self) -> None: self.copySelectedShape() self.pasteSelectedShape() def pasteSelectedShape(self) -> None: self._load_shapes(shapes=self._copied_shapes, replace=False) self._canvas_widgets.canvas.selectShapes(self._copied_shapes) self.setDirty() def copySelectedShape(self) -> None: self._copied_shapes = [ s.copy() for s in self._canvas_widgets.canvas.selectedShapes ] self._actions.paste.setEnabled(len(self._copied_shapes) > 0) def _label_selection_changed(self) -> None: selected_shapes: list[Shape] = [] for item in self._docks.label_list.selectedItems(): selected_shapes.append(item.shape()) if selected_shapes: self._canvas_widgets.canvas.selectShapes(selected_shapes) else: if self._canvas_widgets.canvas.deSelectShape(): self._canvas_widgets.canvas.update() def labelItemChanged(self, item: LabelListWidgetItem) -> None: shape = item.shape() self._canvas_widgets.canvas.setShapeVisible( shape, item.checkState() == Qt.Checked ) def labelOrderChanged(self) -> None: self.setDirty() self._canvas_widgets.canvas.loadShapes( [item.shape() for item in self._docks.label_list] ) # Callback functions: def newShape(self) -> None: """Pop-up and give focus to the label editor. position MUST be in global coordinates. """ items = self._docks.unique_label_list.selectedItems() text = None if items: text = items[0].data(Qt.UserRole) flags = {} group_id = None description = "" if self._config["display_label_popup"] or not text: previous_text = self._label_dialog.edit.text() text, flags, group_id, description = self._label_dialog.popUp(text) if not text: self._label_dialog.edit.setText(previous_text) if text and not self.validateLabel(text): self.errorMessage( self.tr("Invalid label"), self.tr("Invalid label '{}' with validation type '{}'").format( text, self._config["validate_label"] ), ) text = "" if text: self._docks.label_list.clearSelection() shape = self._canvas_widgets.canvas.setLastLabel(text, flags) shape.group_id = group_id shape.description = description self.addLabel(shape) self._actions.edit_mode.setEnabled(True) self._actions.undo_last_point.setEnabled(False) self._actions.undo.setEnabled(True) self.setDirty() else: self._canvas_widgets.canvas.undoLastLine() self._canvas_widgets.canvas.shapesBackups.pop() def scrollRequest(self, delta: int, orientation: Qt.Orientation) -> None: units = -delta * 0.1 # natural scroll bar = self._canvas_widgets.scroll_bars[orientation] value = bar.value() + bar.singleStep() * units self.setScroll(orientation, value) def setScroll(self, orientation: Qt.Orientation, value: float) -> None: self._canvas_widgets.scroll_bars[orientation].setValue(int(value)) if self._filename is not None: self._scroll_values[orientation][self._filename] = value def _set_zoom(self, value: int, pos: QtCore.QPointF | None = None) -> None: if self._filename is None: logger.warning("filename is None, cannot set zoom") return if pos is None: pos = QtCore.QPointF( self._canvas_widgets.canvas.visibleRegion().boundingRect().center() ) canvas_width_old: int = self._canvas_widgets.canvas.width() self._actions.fit_width.setChecked(self._zoom_mode == _ZoomMode.FIT_WIDTH) self._actions.fit_window.setChecked(self._zoom_mode == _ZoomMode.FIT_WINDOW) self._canvas_widgets.canvas.enableDragging( enabled=value > int(self._scalers[_ZoomMode.FIT_WINDOW]() * 100) ) self._canvas_widgets.zoom_widget.setValue(value) # triggers self._paint_canvas self._zoom_values[self._filename] = (self._zoom_mode, value) canvas_width_new: int = self._canvas_widgets.canvas.width() if canvas_width_old == canvas_width_new: return canvas_scale_factor = canvas_width_new / canvas_width_old x_shift: float = pos.x() * canvas_scale_factor - pos.x() y_shift: float = pos.y() * canvas_scale_factor - pos.y() self.setScroll( Qt.Horizontal, self._canvas_widgets.scroll_bars[Qt.Horizontal].value() + x_shift, ) self.setScroll( Qt.Vertical, self._canvas_widgets.scroll_bars[Qt.Vertical].value() + y_shift, ) def _set_zoom_to_original(self): self._zoom_mode = _ZoomMode.MANUAL_ZOOM self._set_zoom(value=100) def _add_zoom(self, increment: float, pos: QtCore.QPointF | None = None) -> None: zoom_value: int if increment > 1: zoom_value = math.ceil(self._canvas_widgets.zoom_widget.value() * increment) else: zoom_value = math.floor( self._canvas_widgets.zoom_widget.value() * increment ) self._zoom_mode = _ZoomMode.MANUAL_ZOOM self._set_zoom(value=zoom_value, pos=pos) def _zoom_requested(self, delta: int, pos: QtCore.QPointF) -> None: self._add_zoom(increment=1.1 if delta > 0 else 0.9, pos=pos) def setFitWindow(self, value=True): if value: self._actions.fit_width.setChecked(False) self._zoom_mode = _ZoomMode.FIT_WINDOW if value else _ZoomMode.MANUAL_ZOOM self._adjust_scale() def setFitWidth(self, value=True): if value: self._actions.fit_window.setChecked(False) self._zoom_mode = _ZoomMode.FIT_WIDTH if value else _ZoomMode.MANUAL_ZOOM self._adjust_scale() def enableKeepPrevScale(self, enabled): self._config["keep_prev_scale"] = enabled self._actions.keep_prev_scale.setChecked(enabled) def onNewBrightnessContrast(self, qimage): self._canvas_widgets.canvas.loadPixmap( QtGui.QPixmap.fromImage(qimage), clear_shapes=False ) def brightnessContrast(self, value: bool, is_initial_load: bool = False): if self._filename is None: logger.warning("filename is None, cannot set brightness/contrast") return brightness: int | None contrast: int | None brightness, contrast = self._brightness_contrast_values.get( self._filename, (None, None) ) if is_initial_load: prev_filename: str = self._recent_files[0] if self._recent_files else "" if self._config["keep_prev_brightness_contrast"] and prev_filename: brightness, contrast = self._brightness_contrast_values.get( prev_filename, (None, None) ) if brightness is None and contrast is None: return logger.debug( "Opening brightness/contrast dialog with brightness={}, contrast={}", brightness, contrast, ) dialog = BrightnessContrastDialog( utils.img_data_to_pil(self.imageData), self.onNewBrightnessContrast, parent=self, ) if brightness is not None: dialog.slider_brightness.setValue(brightness) if contrast is not None: dialog.slider_contrast.setValue(contrast) if is_initial_load: dialog.onNewValue(None) else: dialog.exec_() brightness = dialog.slider_brightness.value() contrast = dialog.slider_contrast.value() self._brightness_contrast_values[self._filename] = (brightness, contrast) logger.debug( "Updated states for {}: brightness={}, contrast={}", self._filename, brightness, contrast, ) def toggleShapes(self, value): flag = value for item in self._docks.label_list: if value is None: flag = item.checkState() == Qt.Unchecked item.setCheckState(Qt.Checked if flag else Qt.Unchecked) def _load_file(self, filename=None): """Load the specified file, or the last opened file if None.""" # changing fileListWidget loads file if filename in self.imageList and ( self._docks.file_list.currentRow() != self.imageList.index(filename) ): self._docks.file_list.setCurrentRow(self.imageList.index(filename)) self._docks.file_list.repaint() return prev_shapes: list[Shape] = ( self._canvas_widgets.canvas.shapes if self._config["keep_prev"] or QtWidgets.QApplication.keyboardModifiers() == (Qt.ControlModifier | Qt.ShiftModifier) else [] ) self.resetState() self._canvas_widgets.canvas.setEnabled(False) if filename is None: filename = self.settings.value("filename", "") filename = str(filename) if not QtCore.QFile.exists(filename): self.errorMessage( self.tr("Error opening file"), self.tr("No such file: %s") % filename, ) return False # assumes same name, but json extension self.show_status_message(self.tr("Loading %s...") % osp.basename(str(filename))) t0_load_file = time.time() label_file = f"{osp.splitext(filename)[0]}.json" if self._output_dir: label_file_without_path = osp.basename(label_file) label_file = osp.join(self._output_dir, label_file_without_path) if QtCore.QFile.exists(label_file) and LabelFile.is_label_file(label_file): try: self._label_file = LabelFile(label_file) except LabelFileError as e: self.errorMessage( self.tr("Error opening file"), self.tr( "%s
" "Make sure %s is a valid label file.
" ) % (e, label_file), ) self.show_status_message(self.tr("Error reading %s") % label_file) return False assert self._label_file is not None self.imageData = self._label_file.imageData assert self._label_file.imagePath self._image_path = osp.join( osp.dirname(label_file), self._label_file.imagePath, ) self._other_data = self._label_file.otherData else: try: self.imageData = LabelFile.load_image_file(filename) except OSError as e: self.errorMessage( self.tr("Error opening file"), self.tr( "%s
" "Make sure %s is a valid image file.
" ) % (e, filename), ) self.show_status_message(self.tr("Error reading %s") % filename) return False if self.imageData: self._image_path = filename self._label_file = None assert self.imageData is not None t0 = time.time() image = QtGui.QImage.fromData(self.imageData) logger.debug("Created QImage in {:.0f}ms", (time.time() - t0) * 1000) if image.isNull(): formats = [ f"*.{fmt.data().decode()}" for fmt in QtGui.QImageReader.supportedImageFormats() ] self.errorMessage( self.tr("Error opening file"), self.tr( "Make sure {0} is a valid image file.
"
"Supported image formats: {1}
{title}
{message}" ) def currentPath(self) -> str: return osp.dirname(str(self._filename)) if self._filename else "." def toggleKeepPrevMode(self) -> None: self._config["keep_prev"] = not self._config["keep_prev"] def removeSelectedPoint(self) -> None: self._canvas_widgets.canvas.removeSelectedPoint() self._canvas_widgets.canvas.update() if ( self._canvas_widgets.canvas.hShape and not self._canvas_widgets.canvas.hShape.points ): self._canvas_widgets.canvas.deleteShape(self._canvas_widgets.canvas.hShape) self.remLabels([self._canvas_widgets.canvas.hShape]) if self.noShapes(): for action in self._actions.on_shapes_present: action.setEnabled(False) self.setDirty() def deleteSelectedShape(self) -> None: yes, no = QtWidgets.QMessageBox.Yes, QtWidgets.QMessageBox.No msg = self.tr( "You are about to permanently delete {} shapes, proceed anyway?" ).format(len(self._canvas_widgets.canvas.selectedShapes)) if yes == QtWidgets.QMessageBox.warning( self, self.tr("Attention"), msg, yes | no, yes ): self.remLabels(self._canvas_widgets.canvas.deleteSelected()) self.setDirty() if self.noShapes(): for action in self._actions.on_shapes_present: action.setEnabled(False) def copyShape(self) -> None: self._canvas_widgets.canvas.endMove(copy=True) for shape in self._canvas_widgets.canvas.selectedShapes: self.addLabel(shape) self._docks.label_list.clearSelection() self.setDirty() def moveShape(self) -> None: self._canvas_widgets.canvas.endMove(copy=False) self.setDirty() def _open_dir_with_dialog(self, _value: bool = False) -> None: if not self._can_continue(): return defaultOpenDirPath: str if self._prev_opened_dir and osp.exists(self._prev_opened_dir): defaultOpenDirPath = self._prev_opened_dir else: defaultOpenDirPath = osp.dirname(self._filename) if self._filename else "." targetDirPath = str( QtWidgets.QFileDialog.getExistingDirectory( self, self.tr("%s - Open Directory") % __appname__, defaultOpenDirPath, QtWidgets.QFileDialog.ShowDirsOnly | QtWidgets.QFileDialog.DontResolveSymlinks, ) ) self._import_images_from_dir(root_dir=targetDirPath) self._open_next_image() @property def imageList(self) -> list[str]: lst = [] for i in range(self._docks.file_list.count()): item = self._docks.file_list.item(i) assert item lst.append(item.text()) return lst def importDroppedImageFiles(self, imageFiles): extensions = [ f".{fmt.data().decode().lower()}" for fmt in QtGui.QImageReader.supportedImageFormats() ] self._filename = None for file in imageFiles: if file in self.imageList or not file.lower().endswith(tuple(extensions)): continue label_file = f"{osp.splitext(file)[0]}.json" if self._output_dir: label_file_without_path = osp.basename(label_file) label_file = osp.join(self._output_dir, label_file_without_path) item = QtWidgets.QListWidgetItem(file) item.setFlags(Qt.ItemIsEnabled | Qt.ItemIsSelectable) if QtCore.QFile.exists(label_file) and LabelFile.is_label_file(label_file): item.setCheckState(Qt.Checked) else: item.setCheckState(Qt.Unchecked) self._docks.file_list.addItem(item) if len(self.imageList) > 1: self._actions.open_next_img.setEnabled(True) self._actions.open_prev_img.setEnabled(True) self._open_next_image() def _import_images_from_dir( self, root_dir: str | None, pattern: str | None = None ) -> None: self._actions.open_next_img.setEnabled(True) self._actions.open_prev_img.setEnabled(True) if not self._can_continue() or not root_dir: return self._prev_opened_dir = root_dir self._filename = None self._docks.file_list.clear() filenames = _scan_image_files(root_dir=root_dir) if pattern: try: filenames = [f for f in filenames if re.search(pattern, f)] except re.error: pass for filename in filenames: label_file = f"{osp.splitext(filename)[0]}.json" if self._output_dir: label_file_without_path = osp.basename(label_file) label_file = osp.join(self._output_dir, label_file_without_path) item = QtWidgets.QListWidgetItem(filename) item.setFlags(Qt.ItemIsEnabled | Qt.ItemIsSelectable) if QtCore.QFile.exists(label_file) and LabelFile.is_label_file(label_file): item.setCheckState(Qt.Checked) else: item.setCheckState(Qt.Unchecked) self._docks.file_list.addItem(item) def _update_status_stats(self, mouse_pos: QtCore.QPointF) -> None: stats: list[str] = [] stats.append(f"mode={self._canvas_widgets.canvas.mode.name}") stats.append(f"x={mouse_pos.x():6.1f}, y={mouse_pos.y():6.1f}") self._status_bar.stats.setText(" | ".join(stats)) def _scan_image_files(root_dir: str) -> list[str]: extensions: list[str] = [ f".{fmt.data().decode().lower()}" for fmt in QtGui.QImageReader.supportedImageFormats() ] images: list[str] = [] for root, dirs, files in os.walk(root_dir): for file in files: if file.lower().endswith(tuple(extensions)): relativePath = os.path.normpath(osp.join(root, file)) images.append(relativePath) logger.debug("found {:d} images in {!r}", len(images), root_dir) return natsort.os_sorted(images) ================================================ FILE: labelme/config/__init__.py ================================================ import os.path as osp import re from pathlib import Path import yaml from loguru import logger here = osp.dirname(osp.abspath(__file__)) def _update_dict(target_dict, new_dict, validate_item=None): for key, value in new_dict.items(): if validate_item: validate_item(key, value) if key not in target_dict: raise ValueError(f"Unexpected key in config: {key}") if isinstance(target_dict[key], dict) and isinstance(value, dict): _update_dict(target_dict[key], value, validate_item=validate_item) else: target_dict[key] = value def _validate_config_item(key, value): if key == "validate_label" and value not in [None, "exact"]: raise ValueError(f"Unexpected value for config key 'validate_label': {value}") if key == "shape_color" and value not in [None, "auto", "manual"]: raise ValueError(f"Unexpected value for config key 'shape_color': {value}") if key == "labels" and value is not None and len(value) != len(set(value)): raise ValueError(f"Duplicates are detected for config key 'labels': {value}") def _migrate_config_from_file(config_from_yaml: dict) -> None: keep_prev_brightness: bool = config_from_yaml.pop("keep_prev_brightness", False) keep_prev_contrast: bool = config_from_yaml.pop("keep_prev_contrast", False) if keep_prev_brightness or keep_prev_contrast: logger.info( "Migrating old config: keep_prev_brightness={} or keep_prev_contrast={} " "-> keep_prev_brightness_contrast=True", keep_prev_brightness, keep_prev_contrast, ) config_from_yaml["keep_prev_brightness_contrast"] = True if "store_data" in config_from_yaml: logger.info("Migrating old config: store_data -> with_image_data") config_from_yaml["with_image_data"] = config_from_yaml.pop("store_data") if config_from_yaml.get("shortcuts", {}).pop("add_point_to_edge", None): logger.info("Migrating old config: removing shortcuts.add_point_to_edge") if (model_name := config_from_yaml.get("ai", {}).get("default")) and ( m := re.match(r"^SegmentAnything \((.*)\)$", model_name) ): model_name_new: str = f"Sam ({m.group(1)})" logger.info( "Migrating old config: ai.default={!r} -> ai.default={!r}", model_name, model_name_new, ) config_from_yaml["ai"]["default"] = model_name_new # Migrate polygon shortcut keys to shape _POLYGON_TO_SHAPE_RENAMES = { "edit_polygon": "edit_shape", "delete_polygon": "delete_shape", "duplicate_polygon": "duplicate_shape", "copy_polygon": "copy_shape", "paste_polygon": "paste_shape", "show_all_polygons": "show_all_shapes", "hide_all_polygons": "hide_all_shapes", "toggle_all_polygons": "toggle_all_shapes", } shortcuts = config_from_yaml.get("shortcuts", {}) for old_key, new_key in _POLYGON_TO_SHAPE_RENAMES.items(): if old_key in shortcuts and new_key not in shortcuts: logger.info( "Migrating old config: shortcuts.{} -> shortcuts.{}", old_key, new_key, ) shortcuts[new_key] = shortcuts.pop(old_key) def get_user_config_file(create_if_missing: bool = True) -> str: user_config_file: str = osp.join(osp.expanduser("~"), ".labelmerc") if not osp.exists(user_config_file) and create_if_missing: try: with open(user_config_file, "w") as f: f.write( "# Labelme config file.\n" "# Only add settings you want to override.\n" "# For all available options and defaults, see:\n" "# https://github.com/wkentaro/labelme/blob/main/labelme/config/default_config.yaml\n" "#\n" "# Example:\n" "# with_image_data: true\n" "# auto_save: false\n" "# labels: [cat, dog]\n" ) except Exception: logger.warning("Failed to save config: {!r}", user_config_file) return user_config_file def load_config(config_file: Path | None, config_overrides: dict) -> dict: config: dict with open(osp.join(here, "default_config.yaml")) as f: config = yaml.safe_load(f) if config_file is not None: with open(config_file) as f: config_from_yaml = yaml.safe_load(f) if isinstance(config_from_yaml, dict): _migrate_config_from_file(config_from_yaml=config_from_yaml) _update_dict(config, config_from_yaml, validate_item=_validate_config_item) _update_dict(config, config_overrides, validate_item=_validate_config_item) if not config["labels"] and config["validate_label"]: raise ValueError("labels must be specified when validate_label is enabled") return config ================================================ FILE: labelme/config/default_config.yaml ================================================ auto_save: true display_label_popup: true with_image_data: false keep_prev: false keep_prev_scale: false keep_prev_brightness_contrast: false logger_level: info flags: null label_flags: null labels: null file_search: null sort_labels: true validate_label: null default_shape_color: [0, 255, 0] shape_color: auto # null, 'auto', 'manual' shift_auto_shape_color: 0 label_colors: null shape: # drawing line_color: [0, 255, 0, 128] fill_color: [0, 0, 0, 64] vertex_fill_color: [0, 255, 0, 255] # selecting / hovering select_line_color: [255, 255, 255, 255] select_fill_color: [0, 255, 0, 64] hvertex_fill_color: [255, 255, 255, 255] point_size: 8 ai: default: 'Sam2 (balanced)' # main flag_dock: show: true closable: true movable: true floatable: true label_dock: show: true closable: true movable: true floatable: true shape_dock: show: true closable: true movable: true floatable: true file_dock: show: true closable: true movable: true floatable: true # label_dialog show_label_text_field: true label_completion: startswith fit_to_content: column: true row: false # canvas epsilon: 10.0 canvas: fill_drawing: true # None: do nothing # close: close polygon double_click: close # The max number of edits we can undo num_backups: 10 # show crosshair crosshair: polygon: false rectangle: true circle: false line: false point: false linestrip: false ai_polygon: false ai_mask: false shortcuts: close: Ctrl+W open: Ctrl+O open_dir: Ctrl+U quit: Ctrl+Q save: Ctrl+S save_as: Ctrl+Shift+S save_to: null delete_file: Ctrl+Delete open_next: [D, Ctrl+Shift+D] open_prev: [A, Ctrl+Shift+A] zoom_in: [Ctrl++, Ctrl+=] zoom_out: Ctrl+- zoom_to_original: Ctrl+0 fit_window: Ctrl+F fit_width: Ctrl+Shift+F create_polygon: Ctrl+N create_rectangle: Ctrl+R create_circle: null create_line: null create_point: null create_linestrip: null edit_shape: Ctrl+J delete_shape: Delete duplicate_shape: Ctrl+D copy_shape: Ctrl+C paste_shape: Ctrl+V undo: Ctrl+Z undo_last_point: Ctrl+Z edit_label: Ctrl+E toggle_keep_prev_mode: Ctrl+P remove_selected_point: [Meta+H, Backspace] show_all_shapes: null hide_all_shapes: null toggle_all_shapes: T ================================================ FILE: labelme/shape.py ================================================ from __future__ import annotations import copy import numpy as np import numpy.typing as npt import skimage.measure from loguru import logger from PyQt5 import QtCore from PyQt5 import QtGui import labelme.utils class Shape: # Handle point styles: square or round P_SQUARE = 0 P_ROUND = 1 # Vertex interaction modes MOVE_VERTEX = 0 NEAR_VERTEX = 1 # Width of the shape outline pen PEN_WIDTH = 2 # The following class variables influence the drawing of all shape objects. line_color: QtGui.QColor = QtGui.QColor(0, 255, 0, 128) fill_color: QtGui.QColor = QtGui.QColor(0, 0, 0, 64) vertex_fill_color: QtGui.QColor = QtGui.QColor(0, 255, 0, 255) select_line_color: QtGui.QColor = QtGui.QColor(255, 255, 255, 255) select_fill_color: QtGui.QColor = QtGui.QColor(0, 255, 0, 64) hvertex_fill_color: QtGui.QColor = QtGui.QColor(255, 255, 255, 255) # Default handle style, size, and zoom scale point_type = P_ROUND point_size = 8 scale = 1.0 _current_vertex_fill_color: QtGui.QColor def __init__( self, label: str | None = None, line_color: QtGui.QColor | None = None, shape_type: str | None = None, flags: dict[str, bool] | None = None, group_id: int | None = None, description: str | None = None, mask: npt.NDArray[np.bool_] | None = None, ): self.label = label self.group_id = group_id self.points = [] self.point_labels = [] self.shape_type = shape_type self._shape_raw = None self._points_raw = [] self._shape_type_raw = None self.fill = False self.selected = False self.flags = flags self.description = description self.other_data = {} self.mask = mask # Highlight state: which vertex is highlighted and how it looks self._highlightIndex = None self._highlightMode = self.NEAR_VERTEX self._highlightSettings = { self.NEAR_VERTEX: (4, self.P_ROUND), self.MOVE_VERTEX: (1.5, self.P_SQUARE), } self._closed = False if line_color is not None: # Per-instance line color override (used for the pending line). self.line_color = line_color def _scale_point(self, point: QtCore.QPointF) -> QtCore.QPointF: return QtCore.QPointF(point.x() * self.scale, point.y() * self.scale) def setShapeRefined(self, shape_type, points, point_labels, mask=None): self._shape_raw = (self.shape_type, self.points, self.point_labels) self.shape_type = shape_type self.points = points self.point_labels = point_labels self.mask = mask def restoreShapeRaw(self): if self._shape_raw is None: return self.shape_type, self.points, self.point_labels = self._shape_raw self._shape_raw = None @property def shape_type(self): return self._shape_type @shape_type.setter def shape_type(self, value): if value is None: value = "polygon" if value not in [ "polygon", "rectangle", "point", "line", "circle", "linestrip", "points", "mask", ]: raise ValueError(f"Unexpected shape_type: {value}") self._shape_type = value def close(self): self._closed = True def addPoint(self, point, label=1): if self.points and point == self.points[0]: self.close() else: self.points.append(point) self.point_labels.append(label) def canAddPoint(self): return self.shape_type in ["polygon", "linestrip"] def popPoint(self): if self.points: if self.point_labels: self.point_labels.pop() return self.points.pop() return None def insertPoint(self, i, point, label=1): self.points.insert(i, point) self.point_labels.insert(i, label) def canRemovePoint(self) -> bool: if not self.canAddPoint(): return False if self.shape_type == "polygon" and len(self.points) <= 3: return False if self.shape_type == "linestrip" and len(self.points) <= 2: return False return True def removePoint(self, i: int): if not self.canRemovePoint(): logger.warning( "Cannot remove point from: shape_type=%r, len(points)=%d", self.shape_type, len(self.points), ) return self.points.pop(i) self.point_labels.pop(i) def isClosed(self): return self._closed def setOpen(self): self._closed = False def paint(self, painter): if self.mask is None and not self.points: return color = self.select_line_color if self.selected else self.line_color pen = QtGui.QPen(color) # Try using integer sizes for smoother drawing(?) pen.setWidth(self.PEN_WIDTH) painter.setPen(pen) if self.shape_type == "mask" and self.mask is not None: image_to_draw = np.zeros(self.mask.shape + (4,), dtype=np.uint8) fill_color = ( self.select_fill_color.getRgb() if self.selected else self.fill_color.getRgb() ) image_to_draw[self.mask] = fill_color qimage = QtGui.QImage.fromData(labelme.utils.img_arr_to_data(image_to_draw)) qimage = qimage.scaled( qimage.size() * self.scale, QtCore.Qt.IgnoreAspectRatio, QtCore.Qt.SmoothTransformation, ) painter.drawImage(self._scale_point(point=self.points[0]), qimage) line_path = QtGui.QPainterPath() contours = skimage.measure.find_contours(np.pad(self.mask, pad_width=1)) for contour in contours: contour += [self.points[0].y(), self.points[0].x()] line_path.moveTo( self._scale_point(QtCore.QPointF(contour[0, 1], contour[0, 0])) ) for point in contour[1:]: line_path.lineTo( self._scale_point(QtCore.QPointF(point[1], point[0])) ) painter.drawPath(line_path) if self.points: line_path = QtGui.QPainterPath() vrtx_path = QtGui.QPainterPath() negative_vrtx_path = QtGui.QPainterPath() if self.shape_type in ["rectangle", "mask"]: assert len(self.points) in [1, 2] if len(self.points) == 2: rectangle = QtCore.QRectF( self._scale_point(self.points[0]), self._scale_point(self.points[1]), ) line_path.addRect(rectangle) if self.shape_type == "rectangle": for i in range(len(self.points)): self.drawVertex(vrtx_path, i) elif self.shape_type == "circle": assert len(self.points) in [1, 2] if len(self.points) == 2: radius = labelme.utils.distance( self._scale_point(self.points[0] - self.points[1]) ) line_path.addEllipse( self._scale_point(self.points[0]), radius, radius ) for i in range(len(self.points)): self.drawVertex(vrtx_path, i) elif self.shape_type == "linestrip": line_path.moveTo(self._scale_point(self.points[0])) for i, p in enumerate(self.points): line_path.lineTo(self._scale_point(p)) self.drawVertex(vrtx_path, i) elif self.shape_type == "points": assert len(self.points) == len(self.point_labels) for i, point_label in enumerate(self.point_labels): if point_label == 1: self.drawVertex(vrtx_path, i) else: self.drawVertex(negative_vrtx_path, i) else: line_path.moveTo(self._scale_point(self.points[0])) # Uncommenting the following line will draw 2 paths # for the 1st vertex, and make it non-filled, which # may be desirable. # self.drawVertex(vrtx_path, 0) for i, p in enumerate(self.points): line_path.lineTo(self._scale_point(p)) self.drawVertex(vrtx_path, i) if self.isClosed(): line_path.lineTo(self._scale_point(self.points[0])) painter.drawPath(line_path) if vrtx_path.length() > 0: painter.drawPath(vrtx_path) painter.fillPath(vrtx_path, self._current_vertex_fill_color) if self.fill and self.shape_type not in [ "line", "linestrip", "points", "mask", ]: color = self.select_fill_color if self.selected else self.fill_color painter.fillPath(line_path, color) pen.setColor(QtGui.QColor(255, 0, 0, 255)) painter.setPen(pen) painter.drawPath(negative_vrtx_path) painter.fillPath(negative_vrtx_path, QtGui.QColor(255, 0, 0, 255)) def drawVertex(self, path, i): d = self.point_size shape = self.point_type point = self._scale_point(self.points[i]) if i == self._highlightIndex: size, shape = self._highlightSettings[self._highlightMode] d *= size # type: ignore[assignment] if self._highlightIndex is not None: self._current_vertex_fill_color = self.hvertex_fill_color else: self._current_vertex_fill_color = self.vertex_fill_color if shape == self.P_SQUARE: path.addRect(point.x() - d / 2, point.y() - d / 2, d, d) elif shape == self.P_ROUND: path.addEllipse(point, d / 2.0, d / 2.0) else: assert False, "unsupported vertex shape" def nearestVertex(self, point: QtCore.QPointF, epsilon: float) -> int | None: min_distance = float("inf") min_i = None point = self._scale_point(point) for i, p in enumerate(self.points): p = self._scale_point(p) dist = labelme.utils.distance(p - point) if dist <= epsilon and dist < min_distance: min_distance = dist min_i = i return min_i def nearestEdge(self, point: QtCore.QPointF, epsilon: float) -> int | None: min_distance = float("inf") post_i = None point = self._scale_point(point) for i in range(len(self.points)): start = self._scale_point(self.points[i - 1]) end = self._scale_point(self.points[i]) line = [start, end] dist = labelme.utils.distancetoline(point, line) if dist <= epsilon and dist < min_distance: min_distance = dist post_i = i return post_i def containsPoint(self, point: QtCore.QPointF) -> bool: if self.shape_type in ["line", "linestrip", "points"]: return False if self.shape_type == "point": if not self.points: return False return labelme.utils.distance(point - self.points[0]) <= self.point_size / 2 if self.mask is not None: raw_y = int(round(point.y() - self.points[0].y())) raw_x = int(round(point.x() - self.points[0].x())) if ( raw_y < 0 or raw_y >= self.mask.shape[0] or raw_x < 0 or raw_x >= self.mask.shape[1] ): return False return bool(self.mask[raw_y, raw_x]) return self.makePath().contains(point) def makePath(self): if self.shape_type in ["rectangle", "mask"]: path = QtGui.QPainterPath() if len(self.points) == 2: path.addRect(QtCore.QRectF(self.points[0], self.points[1])) elif self.shape_type == "circle": path = QtGui.QPainterPath() if len(self.points) == 2: raidus = labelme.utils.distance(self.points[0] - self.points[1]) path.addEllipse(self.points[0], raidus, raidus) else: path = QtGui.QPainterPath(self.points[0]) for p in self.points[1:]: path.lineTo(p) return path def boundingRect(self): return self.makePath().boundingRect() def moveBy(self, offset): self.points = [p + offset for p in self.points] def moveVertex(self, i: int, pos: QtCore.QPointF) -> None: self.points[i] = pos def highlightVertex(self, i: int, action: int) -> None: self._highlightIndex = i self._highlightMode = action def highlightClear(self) -> None: self._highlightIndex = None def copy(self): return copy.deepcopy(self) def __len__(self): return len(self.points) def __getitem__(self, key): return self.points[key] def __setitem__(self, key, value): self.points[key] = value ================================================ FILE: labelme/testing.py ================================================ import json import os.path as osp import imgviz import labelme.utils def assert_labelfile_sanity(filename): assert osp.exists(filename) data = json.load(open(filename)) assert "imagePath" in data imageData = data.get("imageData", None) if imageData is None: parent_dir = osp.dirname(filename) img_file = osp.join(parent_dir, data["imagePath"]) assert osp.exists(img_file) img = imgviz.io.imread(img_file) else: img = labelme.utils.img_b64_to_arr(imageData) H, W = img.shape[:2] assert H == data["imageHeight"] assert W == data["imageWidth"] assert "shapes" in data for shape in data["shapes"]: assert "label" in shape assert "points" in shape for x, y in shape["points"]: assert 0 <= x <= W assert 0 <= y <= H ================================================ FILE: labelme/translate/de_DE.ts ================================================