[ { "path": ".gitignore", "content": ".DS_Store\n/.build\n/Packages\n/*.xcodeproj\nxcuserdata/\nDerivedData/\n.swiftpm/xcode/package.xcworkspace/contents.xcworkspacedata\n" }, { "path": ".swiftpm/xcode/package.xcworkspace/xcshareddata/IDEWorkspaceChecks.plist", "content": "\n\n\n\n\tIDEDidComputeMac32BitWarning\n\t\n\n\n" }, { "path": "LICENSE", "content": "MIT License\n\nCopyright (c) 2021 MLBoy\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE.\n" }, { "path": "Package.swift", "content": "// swift-tools-version:5.5\n// The swift-tools-version declares the minimum version of Swift required to build this package.\n\nimport PackageDescription\n\nlet package = Package(\n name: \"SemanticImage\",\n platforms: [\n .iOS(.v14)\n ],\n products: [\n // Products define the executables and libraries a package produces, and make them visible to other packages.\n .library(\n name: \"SemanticImage\",\n targets: [\"SemanticImage\"]),\n ],\n dependencies: [\n // Dependencies declare other packages that this package depends on.\n // .package(url: /* package url */, from: \"1.0.0\"),\n ],\n targets: [\n // Targets are the basic building blocks of a package. A target can define a module or a test suite.\n // Targets can depend on other targets in this package, and on products in packages this package depends on.\n .target(\n name: \"SemanticImage\",\n dependencies: [],\n resources: [.process(\"segmentation.mlmodelc\")]),\n .testTarget(\n name: \"SemanticImageTests\",\n dependencies: [\"SemanticImage\"]),\n ]\n)\n" }, { "path": "README.md", "content": "# SemanticImage\n\n**Plug-and-play image/video filters built on Vision + CoreML + CoreImage ā€” for iOS. Person segmentation, background blur, depth-based bokeh, style transfer, super-resolution.**\n\nBy [Daisuke Majima](https://john-rocky.github.io). Maintainer of [CoreML-Models](https://github.com/john-rocky/CoreML-Models) (1,749ā˜…).\n\nThese filters power [Blur. on the App Store](https://apps.apple.com/us/developer/daisuke-majima/id1350309854) and similar shipping iOS apps.\n\nšŸ’¼ **Open to Staff / Senior iOS / Mobile ML roles** ā€” [john-rocky.github.io](https://john-rocky.github.io)\n\n---\n\n# SemanticImage\n\nA collection of easy-to-use image / video filters.\n\n# How to use\n\n### Setting Up\n\n1, Add SemanticImage to your project as Swift Package with Swift Package Manager. \n Or just drag SemanticImage.swift to your project.\n\n2, Import and initialize SemanticImage\n\n```swift\nimport SemanticImage\n```\n\n```swift\nlet semanticImage = SemanticImage()\n```\n\n**Requires iOS 14 or above**\n\n# Filter Collection\n\n## Image\n\n### Get Person Mask\n\n \n\n\n```swift\nlet maskImage:UIImage? = semanticImage.personMaskImage(uiImage: yourUIImage)\n```\n\n### Swap the background of a person\n\n \n\n```swift\nlet swappedImage:UIImage? = semanticImage.swapBackgroundOfPerson(personUIImage: yourUIImage, backgroundUIImage: yourBackgroundUIImage)\n```\n\n### Blur the backgrond of a person\n\n \n\n```swift\nlet blurredPersonImage:UIImage? = semanticImage.personBlur(uiImage:UIImage, intensity:Float)\n// Blur intensity: 0~100 \n```\n\n### Get a prominent object mask\n\n \n\n```swift\nlet prominentMaskImage:UIImage? = semanticImage.saliencyMask(uiImage:image)\n```\n\n### Swap the background of the prominent object\n\n \n\n```swift\nlet backgroundSwapImage:UIImage? = semanticImage.saliencyBlend(objectUIImage: image, backgroundUIImage: bgImage)\n```\n\n### Crop a face rectangle\n\n \n\n```swift\nlet faceImage:UIImage? = semanticImage.faceRectangle(uiImage: image)\n```\n\n### Crop a body rectangle\n\n \n\n```swift\nlet bodyImage:UIImage? = semanticImage.humanRectangle(uiImage: image)\n```\n\n### Crop face rectangles\n\n \n\n```swift\nlet faceImages:[UIImage] = semanticImage.faceRectangles(uiImage: image)\n```\n\n### Crop body rectangles\n\n \n\n```swift\nlet bodyImages:[UIImage] = semanticImage.humanRectangles(uiImage: image)\n```\n\n### Crop an animal(Cat/Dog) rectangle\n\n \n\n```swift\nlet animalImage:UIImage? = semanticImage.animalRectangle(uiImage: image)\n```\n\n### Crop multiple animal(Cat/Dog) rectangles\n\n \n\n```swift\nlet animalImages:[UIImage] = semanticImage.animalRectangles(uiImage: image)\n```\n\n### Crop and warp document\n\n \n\n```swift\nlet documentImage:UIImage? = semanticImage.getDocumentImage(image: image)\n```\n\n## Video\n\n### Apply CIFilter to Video\n\n\n\n```swift\nguard let ciFilter = CIFilter(name: \"CIEdgeWork\", parameters: [kCIInputRadiusKey:3.0]) else { return }\nsemanticImage.ciFilterVideo(videoURL: url, ciFilter: ciFilter, { err, processedURL in\n // Handle processedURL in here.\n})\n// This process takes about the same time as the video playback time.\n```\n\n### Add virtual background of the person video\n\n\n\n```swift\nsemanticImage.swapBackgroundOfPersonVideo(videoURL: url, backgroundUIImage: uiImage, { err, processedURL in\n // Handle processedURL in here.\n})\n // This process takes about the same time as the video playback time.\n```\n\n### Add virtual background of the salient object video\n\n\n\n```swift\nsemanticImage.swapBGOfSalientObjectVideo(videoURL: url, backgroundUIImage: uiImage, { err, processedURL in\n // Handle processedURL in here.\n})\n // This process takes about the same time as the video playback time.\n```\n\n### Process video\n\n\n\n```swift\nsemanticImage.applyProcessingOnVideo(videoURL: url, { ciImage in\n // Write the processing of ciImage (i.e. video frame) here.\n return newImage\n}, { err, editedURL in\n // The processed video URL is returned\n})\n```\n\n# Author\n\nDaisuke Majima\n\nFreelance iOS programmer from Japan.\n\nPROFILES:\n\nWORKS:\n\nBLOGS: Medium\n\nCONTACTS: rockyshikoku@gmail.com\n" }, { "path": "Sources/SemanticImage/SemanticImage.swift", "content": "import Foundation\nimport Vision\nimport UIKit\nimport AVKit\n\npublic class SemanticImage {\n \n public init() {\n }\n \n @available(iOS 15.0, *)\n lazy var personSegmentationRequest = VNGeneratePersonSegmentationRequest()\n lazy var faceRectangleRequest = VNDetectFaceRectanglesRequest()\n lazy var humanRectanglesRequest:VNDetectHumanRectanglesRequest = {\n let request = VNDetectHumanRectanglesRequest()\n if #available(iOS 15.0, *) {\n request.upperBodyOnly = false\n }\n return request\n }()\n lazy var animalRequest = VNRecognizeAnimalsRequest()\n\n lazy var segmentationRequest:VNCoreMLRequest? = {\n let url = try? Bundle.module.url(forResource: \"segmentation\", withExtension: \"mlmodelc\")\n let mlModel = try! MLModel(contentsOf: url!, configuration: MLModelConfiguration())\n guard let model = try? VNCoreMLModel(for: mlModel) else { return nil }\n let request = VNCoreMLRequest(model: model)\n request.imageCropAndScaleOption = .scaleFill\n return request\n }()\n lazy var rectangleRequest = VNDetectRectanglesRequest()\n\n let ciContext = CIContext()\n \n public func getDocumentImage(image:UIImage) -> UIImage? {\n let newImage = getCorrectOrientationUIImage(uiImage:image)\n let ciImage = CIImage(image: newImage)!\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n try! handler.perform([rectangleRequest])\n guard let result = rectangleRequest.results?.first else { return nil }\n \n let topLeft = CGPoint(x: result.topLeft.x, y: 1-result.topLeft.y)\n let topRight = CGPoint(x: result.topRight.x, y: 1-result.topRight.y)\n let bottomLeft = CGPoint(x: result.bottomLeft.x, y: 1-result.bottomLeft.y)\n let bottomRight = CGPoint(x: result.bottomRight.x, y: 1-result.bottomRight.y)\n\n \n let deNormalizedTopLeft = VNImagePointForNormalizedPoint(topLeft, Int(ciImage.extent.width), Int(ciImage.extent.height))\n let deNormalizedTopRight = VNImagePointForNormalizedPoint(topRight, Int(ciImage.extent.width), Int(ciImage.extent.height))\n let deNormalizedBottomLeft = VNImagePointForNormalizedPoint(bottomLeft, Int(ciImage.extent.width), Int(ciImage.extent.height))\n let deNormalizedBottomRight = VNImagePointForNormalizedPoint(bottomRight, Int(ciImage.extent.width), Int(ciImage.extent.height))\n\n let croppedImage = getCroppedImage(image: ciImage, topL: deNormalizedTopLeft, topR: deNormalizedTopRight, botL: deNormalizedBottomLeft, botR: deNormalizedBottomRight)\n let safeCGImage = ciContext.createCGImage(croppedImage, from: croppedImage.extent)\n let croppedUIImage = UIImage(cgImage: safeCGImage!)\n return croppedUIImage\n }\n \n private func getCroppedImage(image: CIImage, topL: CGPoint, topR: CGPoint, botL: CGPoint, botR: CGPoint) -> CIImage {\n let rectCoords = NSMutableDictionary(capacity: 4)\n \n rectCoords[\"inputTopLeft\"] = topL.toVector(image: image)\n rectCoords[\"inputTopRight\"] = topR.toVector(image: image)\n rectCoords[\"inputBottomLeft\"] = botL.toVector(image: image)\n rectCoords[\"inputBottomRight\"] = botR.toVector(image: image)\n \n guard let coords = rectCoords as? [String : Any] else {\n return image\n }\n return image.applyingFilter(\"CIPerspectiveCorrection\", parameters: coords)\n }\n \n // MARK: Segmentation\n \n public func personMaskImage(uiImage:UIImage) -> UIImage? {\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n if #available(iOS 15.0, *) {\n try handler.perform([personSegmentationRequest])\n guard let result = personSegmentationRequest.results?.first\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskCIImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskCIImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n\n guard let safeCGImage = ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskUIImage = UIImage(cgImage: safeCGImage)\n return maskUIImage\n } else {\n guard let segmentationRequest = segmentationRequest else {\n print(\"This func can't be used in this OS version.\"); return nil\n }\n try handler.perform([segmentationRequest])\n guard let result = segmentationRequest.results?.first as? VNPixelBufferObservation\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskCIImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskCIImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n\n guard let safeCGImage = ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskUIImage = UIImage(cgImage: safeCGImage)\n return maskUIImage\n }\n } catch let error {\n print(\"Vision error \\(error)\")\n return nil\n }\n }\n \n \n public func swapBackgroundOfPerson(personUIImage: UIImage, backgroundUIImage: UIImage) -> UIImage? {\n let newPersonUIImage = getCorrectOrientationUIImage(uiImage:personUIImage)\n let newBackgroundUIImage = getCorrectOrientationUIImage(uiImage:backgroundUIImage)\n \n guard let personCIImage = CIImage(image: newPersonUIImage),\n let backgroundCIImage = CIImage(image: newBackgroundUIImage),\n let maskUIImage = personMaskImage(uiImage: newPersonUIImage),\n let maskCIImage = CIImage(image: maskUIImage) else {\n return nil }\n \n let backgroundImageSize = backgroundCIImage.extent\n let originalSize = personCIImage.extent\n var scale:CGFloat = 1\n let widthScale = originalSize.width / backgroundImageSize.width\n let heightScale = originalSize.height / backgroundImageSize.height\n if widthScale > heightScale {\n scale = personCIImage.extent.width / backgroundImageSize.width\n } else {\n scale = personCIImage.extent.height / backgroundImageSize.height\n }\n \n let scaledBG = backgroundCIImage.resize(as: CGSize(width: backgroundCIImage.extent.width*scale, height: backgroundCIImage.extent.height*scale))\n let BGCenter = CGPoint(x: scaledBG.extent.width/2, y: scaledBG.extent.height/2)\n let originalExtent = personCIImage.extent\n let cropRect = CGRect(x: BGCenter.x-(originalExtent.width/2), y: BGCenter.y-(originalExtent.height/2), width: originalExtent.width, height: originalExtent.height)\n let croppedBG = scaledBG.cropped(to: cropRect)\n let translate = CGAffineTransform(translationX: -croppedBG.extent.minX, y: -croppedBG.extent.minY)\n let traslatedBG = croppedBG.transformed(by: translate)\n guard let blended = CIFilter(name: \"CIBlendWithMask\", parameters: [\n kCIInputImageKey: personCIImage,\n kCIInputBackgroundImageKey:traslatedBG,\n kCIInputMaskImageKey:maskCIImage])?.outputImage else { return nil }\n guard let safeCGImage = ciContext.createCGImage(blended, from: blended.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let blendedUIImage = UIImage(cgImage: safeCGImage)\n return blendedUIImage\n }\n \n public func personBlur(uiImage:UIImage, intensity:Float) -> UIImage?{\n let newUIImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let originalCIImage = CIImage(image: newUIImage),\n let maskUIImage = personMaskImage(uiImage: newUIImage),\n let maskCIImage = CIImage(image: maskUIImage) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let safeCropSize = CGRect(x: 0, y: 0, width: originalCIImage.extent.width * 0.999, height: originalCIImage.extent.height * 0.999)\n guard let blurBGCIImage = CIFilter(name: \"CIGaussianBlur\", parameters: [kCIInputImageKey:originalCIImage,\n kCIInputRadiusKey:intensity])?.outputImage?.cropped(to: safeCropSize).resize(as: originalCIImage.extent.size) else { return nil }\n guard let blendedCIImage = CIFilter(name: \"CIBlendWithMask\", parameters: [\n kCIInputImageKey: originalCIImage,\n kCIInputBackgroundImageKey:blurBGCIImage,\n kCIInputMaskImageKey:maskCIImage])?.outputImage,\n let safeCGImage = ciContext.createCGImage(blendedCIImage, from: blendedCIImage.extent)else { print(\"Image processing failed.Please try with another image.\"); return nil }\n \n let final = UIImage(cgImage: safeCGImage)\n return final\n }\n \n // MARK: Saliency\n \n public func saliencyMask(uiImage:UIImage) -> UIImage? {\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage),\n let request = segmentationRequest else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([request])\n guard let result = request.results?.first as? VNPixelBufferObservation\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskCIImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskCIImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n\n guard let safeCGImage = ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskUIImage = UIImage(cgImage: safeCGImage)\n return maskUIImage\n\n } catch let error {\n print(\"Vision error \\(error)\")\n return nil\n }\n }\n \n public func saliencyBlend(objectUIImage:UIImage, backgroundUIImage: UIImage) -> UIImage? {\n let newSaliencyUIImage = getCorrectOrientationUIImage(uiImage:objectUIImage)\n let newBackgroundUIImage = getCorrectOrientationUIImage(uiImage:backgroundUIImage)\n \n guard let personCIImage = CIImage(image: newSaliencyUIImage),\n let backgroundCIImage = CIImage(image: newBackgroundUIImage),\n let maskUIImage = saliencyMask(uiImage: newSaliencyUIImage),\n let maskCIImage = CIImage(image: maskUIImage) else {\n return nil }\n \n let backgroundImageSize = backgroundCIImage.extent\n let originalSize = personCIImage.extent\n var scale:CGFloat = 1\n let widthScale = originalSize.width / backgroundImageSize.width\n let heightScale = originalSize.height / backgroundImageSize.height\n if widthScale > heightScale {\n scale = personCIImage.extent.width / backgroundImageSize.width\n } else {\n scale = personCIImage.extent.height / backgroundImageSize.height\n }\n \n let scaledBG = backgroundCIImage.resize(as: CGSize(width: backgroundCIImage.extent.width*scale, height: backgroundCIImage.extent.height*scale))\n let BGCenter = CGPoint(x: scaledBG.extent.width/2, y: scaledBG.extent.height/2)\n let originalExtent = personCIImage.extent\n let cropRect = CGRect(x: BGCenter.x-(originalExtent.width/2), y: BGCenter.y-(originalExtent.height/2), width: originalExtent.width, height: originalExtent.height)\n let croppedBG = scaledBG.cropped(to: cropRect)\n let translate = CGAffineTransform(translationX: -croppedBG.extent.minX, y: -croppedBG.extent.minY)\n let traslatedBG = croppedBG.transformed(by: translate)\n print(traslatedBG.extent)\n guard let blended = CIFilter(name: \"CIBlendWithMask\", parameters: [\n kCIInputImageKey: personCIImage,\n kCIInputBackgroundImageKey:traslatedBG,\n kCIInputMaskImageKey:maskCIImage])?.outputImage,\n let safeCGImage = ciContext.createCGImage(blended, from: blended.extent) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let blendedUIImage = UIImage(cgImage: safeCGImage)\n return blendedUIImage\n }\n \n // MARK: Rectangle\n \n \n public func faceRectangle(uiImage:UIImage) -> UIImage? {\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([faceRectangleRequest])\n guard let result = faceRectangleRequest.results?.first else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let boundingBox = result.boundingBox\n let faceRect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n var doubleScaleRect = CGRect(x: faceRect.minX - faceRect.width * 0.5, y: faceRect.minY - faceRect.height * 0.5, width: faceRect.width * 2, height: faceRect.height * 2)\n if doubleScaleRect.minX < 0 {\n doubleScaleRect.origin.x = 0\n }\n\n if doubleScaleRect.minY < 0 {\n doubleScaleRect.origin.y = 0\n }\n if doubleScaleRect.maxX > ciImage.extent.maxX {\n doubleScaleRect = CGRect(x: doubleScaleRect.origin.x, y: doubleScaleRect.origin.y, width: ciImage.extent.width - doubleScaleRect.origin.x, height: doubleScaleRect.height)\n }\n if doubleScaleRect.maxY > ciImage.extent.maxY {\n doubleScaleRect = CGRect(x: doubleScaleRect.origin.x, y: doubleScaleRect.origin.y, width: doubleScaleRect.width, height: ciImage.extent.height - doubleScaleRect.origin.y)\n }\n \n let faceImage = ciImage.cropped(to: doubleScaleRect)\n guard let final = ciContext.createCGImage(faceImage, from: faceImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let finalUiimage = UIImage(cgImage: final)\n return finalUiimage\n } catch let error {\n print(\"Vision error \\(error)\")\n return nil\n }\n }\n \n public func faceRectangles(uiImage:UIImage) -> [UIImage] {\n var faceUIImages:[UIImage] = []\n let semaphore = DispatchSemaphore(value: 0)\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([faceRectangleRequest])\n guard let results = faceRectangleRequest.results else { print(\"Image processing failed.Please try with another image.\"); return [] }\n guard !results.isEmpty else { print(\"Image processing failed.Please try with another image.\"); return [] }\n for result in results {\n let boundingBox = result.boundingBox\n let faceRect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n var doubleScaleRect = CGRect(x: faceRect.minX - faceRect.width * 0.5, y: faceRect.minY - faceRect.height * 0.5, width: faceRect.width * 2, height: faceRect.height * 2)\n if doubleScaleRect.minX < 0 {\n doubleScaleRect.origin.x = 0\n }\n\n if doubleScaleRect.minY < 0 {\n doubleScaleRect.origin.y = 0\n }\n if doubleScaleRect.maxX > ciImage.extent.maxX {\n doubleScaleRect = CGRect(x: doubleScaleRect.origin.x, y: doubleScaleRect.origin.y, width: ciImage.extent.width - doubleScaleRect.origin.x, height: doubleScaleRect.height)\n }\n if doubleScaleRect.maxY > ciImage.extent.maxY {\n doubleScaleRect = CGRect(x: doubleScaleRect.origin.x, y: doubleScaleRect.origin.y, width: doubleScaleRect.width, height: ciImage.extent.height - doubleScaleRect.origin.y)\n }\n \n let faceImage = ciImage.cropped(to: doubleScaleRect)\n guard let final = ciContext.createCGImage(faceImage, from: faceImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let finalUiimage = UIImage(cgImage: final)\n faceUIImages.append(finalUiimage)\n if faceUIImages.count == results.count {\n semaphore.signal()\n }\n }\n semaphore.wait()\n return faceUIImages\n } catch let error {\n print(\"Vision error \\(error)\")\n return []\n }\n }\n \n public func humanRectangle(uiImage:UIImage) -> UIImage? {\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([humanRectanglesRequest])\n guard let result = humanRectanglesRequest.results?.first else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let boundingBox = result.boundingBox\n let humanRect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n let humanImage = ciImage.cropped(to: humanRect)\n guard let final = ciContext.createCGImage(humanImage, from: humanImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let finalUiimage = UIImage(cgImage: final)\n return finalUiimage\n } catch let error {\n print(\"Vision error \\(error)\")\n return nil\n }\n }\n \n public func humanRectangles(uiImage:UIImage) -> [UIImage] {\n var bodyUIImages:[UIImage] = []\n let semaphore = DispatchSemaphore(value: 0)\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([humanRectanglesRequest])\n guard let results = humanRectanglesRequest.results else { print(\"Image processing failed.Please try with another image.\"); return [] }\n guard !results.isEmpty else { print(\"Image processing failed.Please try with another image.\"); return [] }\n \n for result in results {\n let boundingBox = result.boundingBox\n let humanRect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n let humanImage = ciImage.cropped(to: humanRect)\n guard let final = ciContext.createCGImage(humanImage, from: humanImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let finalUiimage = UIImage(cgImage: final)\n bodyUIImages.append(finalUiimage)\n if bodyUIImages.count == results.count {\n semaphore.signal()\n }\n }\n semaphore.wait()\n return bodyUIImages\n } catch let error {\n print(\"Vision error \\(error)\")\n return []\n }\n }\n \n public func animalRectangle(uiImage:UIImage) -> UIImage?{\n \n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([animalRequest])\n guard let result = animalRequest.results?.first else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let boundingBox = result.boundingBox\n let rect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n let croppedImage = ciImage.cropped(to: rect)\n guard let final = ciContext.createCGImage(croppedImage, from: croppedImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return nil }\n let finalUiimage = UIImage(cgImage: final)\n return finalUiimage\n } catch let error {\n print(\"Vision error \\(error)\")\n return nil\n }\n }\n \n public func animalRectangles(uiImage:UIImage) -> [UIImage] {\n var animalUIImages:[UIImage] = []\n let semaphore = DispatchSemaphore(value: 0)\n let newImage = getCorrectOrientationUIImage(uiImage:uiImage)\n guard let ciImage = CIImage(image: newImage) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let handler = VNImageRequestHandler(ciImage: ciImage, options: [:])\n do {\n try handler.perform([animalRequest])\n guard let results = animalRequest.results else { print(\"Image processing failed.Please try with another image.\"); return [] }\n guard !results.isEmpty else { print(\"Image processing failed.Please try with another image.\"); return [] }\n \n for result in results {\n let boundingBox = result.boundingBox\n let rect = VNImageRectForNormalizedRect((boundingBox),Int(ciImage.extent.size.width), Int(ciImage.extent.size.height))\n let croppedImage = ciImage.cropped(to: rect)\n guard let final = ciContext.createCGImage(croppedImage, from: croppedImage.extent) else { print(\"Image processing failed.Please try with another image.\"); return [] }\n let finalUiimage = UIImage(cgImage: final)\n animalUIImages.append(finalUiimage)\n if animalUIImages.count == results.count {\n semaphore.signal()\n }\n }\n semaphore.wait()\n return animalUIImages\n } catch let error {\n print(\"Vision error \\(error)\")\n return []\n }\n }\n \n public func swapBGOfSalientObjectVideo(videoURL:URL, backgroundUIImage: UIImage, _ completion: ((_ err: NSError?, _ filteredVideoURL: URL?) -> Void)?) {\n guard let bgCIImage = CIImage(image: backgroundUIImage) else { print(\"background image is nil\") ; return}\n applyProcessingOnVideo(videoURL: videoURL, { ciImage in\n let personCIImage = ciImage\n let backgroundCIImage = bgCIImage\n var maskCIImage:CIImage\n let handler = VNImageRequestHandler(ciImage: personCIImage, options: [:])\n do {\n guard let segmentationRequest = self.segmentationRequest else {\n print(\"This func can't be used in this OS version.\"); return nil\n }\n try handler.perform([segmentationRequest])\n guard let result = segmentationRequest.results?.first as? VNPixelBufferObservation\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n \n guard let safeCGImage = self.ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n maskCIImage = CIImage(cgImage: safeCGImage)\n } catch let error {\n print(\"Vision error \\(error)\")\n return ciImage\n }\n \n let backgroundImageSize = backgroundCIImage.extent\n let originalSize = personCIImage.extent\n var scale:CGFloat = 1\n let widthScale = originalSize.width / backgroundImageSize.width\n let heightScale = originalSize.height / backgroundImageSize.height\n if widthScale > heightScale {\n scale = personCIImage.extent.width / backgroundImageSize.width\n } else {\n scale = personCIImage.extent.height / backgroundImageSize.height\n }\n \n let scaledBG = backgroundCIImage.resize(as: CGSize(width: backgroundCIImage.extent.width*scale, height: backgroundCIImage.extent.height*scale))\n let BGCenter = CGPoint(x: scaledBG.extent.width/2, y: scaledBG.extent.height/2)\n let originalExtent = personCIImage.extent\n let cropRect = CGRect(x: BGCenter.x-(originalExtent.width/2), y: BGCenter.y-(originalExtent.height/2), width: originalExtent.width, height: originalExtent.height)\n let croppedBG = scaledBG.cropped(to: cropRect)\n let translate = CGAffineTransform(translationX: -croppedBG.extent.minX, y: -croppedBG.extent.minY)\n let traslatedBG = croppedBG.transformed(by: translate)\n guard let blended = CIFilter(name: \"CIBlendWithMask\", parameters: [\n kCIInputImageKey: personCIImage,\n kCIInputBackgroundImageKey:traslatedBG,\n kCIInputMaskImageKey:maskCIImage])?.outputImage,\n let safeCGImage = self.ciContext.createCGImage(blended, from: blended.extent) else {return ciImage}\n let outCIImage = CIImage(cgImage: safeCGImage)\n return outCIImage\n } , { err, processedVideoURL in\n guard err == nil else { print(err?.localizedDescription); return }\n completion?(err,processedVideoURL)\n })\n }\n \n public func swapBackgroundOfPersonVideo(videoURL:URL, backgroundUIImage: UIImage, _ completion: ((_ err: NSError?, _ filteredVideoURL: URL?) -> Void)?) {\n guard let bgCIImage = CIImage(image: backgroundUIImage) else { print(\"background image is nil\") ; return}\n applyProcessingOnVideo(videoURL: videoURL, { ciImage in\n let personCIImage = ciImage\n let backgroundCIImage = bgCIImage\n var maskCIImage:CIImage\n let handler = VNImageRequestHandler(ciImage: personCIImage, options: [:])\n do {\n if #available(iOS 15.0, *) {\n try handler.perform([self.personSegmentationRequest])\n guard let result = self.personSegmentationRequest.results?.first\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n\n guard let safeCGImage = self.ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n maskCIImage = CIImage(cgImage: safeCGImage)\n } else {\n guard let segmentationRequest = self.segmentationRequest else {\n print(\"This func can't be used in this OS version.\"); return nil\n }\n try handler.perform([segmentationRequest])\n guard let result = segmentationRequest.results?.first as? VNPixelBufferObservation\n else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n let maskImage = CIImage(cvPixelBuffer: result.pixelBuffer)\n let scaledMask = maskImage.resize(as: CGSize(width: ciImage.extent.width, height: ciImage.extent.height))\n\n guard let safeCGImage = self.ciContext.createCGImage(scaledMask, from: scaledMask.extent) else { print(\"Image processing failed.Please try with another image.\") ; return nil }\n maskCIImage = CIImage(cgImage: safeCGImage)\n }\n } catch let error {\n print(\"Vision error \\(error)\")\n return ciImage\n }\n \n let backgroundImageSize = backgroundCIImage.extent\n let originalSize = personCIImage.extent\n var scale:CGFloat = 1\n let widthScale = originalSize.width / backgroundImageSize.width\n let heightScale = originalSize.height / backgroundImageSize.height\n if widthScale > heightScale {\n scale = personCIImage.extent.width / backgroundImageSize.width\n } else {\n scale = personCIImage.extent.height / backgroundImageSize.height\n }\n \n let scaledBG = backgroundCIImage.resize(as: CGSize(width: backgroundCIImage.extent.width*scale, height: backgroundCIImage.extent.height*scale))\n let BGCenter = CGPoint(x: scaledBG.extent.width/2, y: scaledBG.extent.height/2)\n let originalExtent = personCIImage.extent\n let cropRect = CGRect(x: BGCenter.x-(originalExtent.width/2), y: BGCenter.y-(originalExtent.height/2), width: originalExtent.width, height: originalExtent.height)\n let croppedBG = scaledBG.cropped(to: cropRect)\n let translate = CGAffineTransform(translationX: -croppedBG.extent.minX, y: -croppedBG.extent.minY)\n let traslatedBG = croppedBG.transformed(by: translate)\n guard let blended = CIFilter(name: \"CIBlendWithMask\", parameters: [\n kCIInputImageKey: personCIImage,\n kCIInputBackgroundImageKey:traslatedBG,\n kCIInputMaskImageKey:maskCIImage])?.outputImage,\n let safeCGImage = self.ciContext.createCGImage(blended, from: blended.extent) else {return ciImage}\n let outCIImage = CIImage(cgImage: safeCGImage)\n return outCIImage\n } , { err, processedVideoURL in\n guard err == nil else { print(err?.localizedDescription); return }\n completion?(err,processedVideoURL)\n })\n }\n \n public func ciFilterVideo(videoURL:URL, ciFilter: CIFilter, _ completion: ((_ err: NSError?, _ filteredVideoURL: URL?) -> Void)?) {\n applyProcessingOnVideo(videoURL: videoURL, { ciImage in\n ciFilter.setValue(ciImage, forKey: kCIInputImageKey)\n let outCIImage = ciFilter.outputImage\n return outCIImage\n } , { err, processedVideoURL in\n guard err == nil else { print(err?.localizedDescription as Any); return }\n completion?(err,processedVideoURL)\n })\n }\n \n public func applyProcessingOnVideo(videoURL:URL, _ processingFunction: @escaping ((CIImage) -> CIImage?), _ completion: ((_ err: NSError?, _ processedVideoURL: URL?) -> Void)?) {\n var frame:Int = 0\n var isFrameRotated = false\n let asset = AVURLAsset(url: videoURL)\n let duration = asset.duration.value\n let frameRate = asset.preferredRate\n let totalFrame = frameRate * Float(duration)\n let err: NSError = NSError.init(domain: \"SemanticImage\", code: 999, userInfo: [NSLocalizedDescriptionKey: \"Video Processing Failed\"])\n guard let writingDestinationUrl: URL = try? FileManager.default.url(for: .documentDirectory, in: .userDomainMask, appropriateFor: nil, create: true).appendingPathComponent(\"\\(Date())\" + \".mp4\") else { print(\"nil\"); return}\n\n // setup\n\n guard let reader: AVAssetReader = try? AVAssetReader.init(asset: asset) else {\n completion?(err, nil)\n return\n }\n guard let writer: AVAssetWriter = try? AVAssetWriter(outputURL: writingDestinationUrl, fileType: AVFileType.mov) else {\n completion?(err, nil)\n return\n }\n \n // setup finish closure\n\n var audioFinished: Bool = false\n var videoFinished: Bool = false\n let writtingFinished: (() -> Void) = {\n if audioFinished == true && videoFinished == true {\n writer.finishWriting {\n completion?(nil, writingDestinationUrl)\n }\n reader.cancelReading()\n }\n }\n \n // prepare video reader\n \n let readerVideoOutput: AVAssetReaderTrackOutput = AVAssetReaderTrackOutput(\n track: asset.tracks(withMediaType: AVMediaType.video)[0],\n outputSettings: [\n kCVPixelBufferPixelFormatTypeKey as String: Int(kCVPixelFormatType_420YpCbCr8BiPlanarFullRange),\n ]\n )\n \n reader.add(readerVideoOutput)\n \n // prepare audio reader\n \n var readerAudioOutput: AVAssetReaderTrackOutput!\n if asset.tracks(withMediaType: AVMediaType.audio).count <= 0 {\n audioFinished = true\n } else {\n readerAudioOutput = AVAssetReaderTrackOutput.init(\n track: asset.tracks(withMediaType: AVMediaType.audio)[0],\n outputSettings: [\n AVSampleRateKey: 44100,\n AVFormatIDKey: kAudioFormatLinearPCM,\n ]\n )\n if reader.canAdd(readerAudioOutput) {\n reader.add(readerAudioOutput)\n } else {\n print(\"Cannot add audio output reader\")\n audioFinished = true\n }\n }\n \n // prepare video input\n \n let transform = asset.tracks(withMediaType: AVMediaType.video)[0].preferredTransform\n let radians = atan2(transform.b, transform.a)\n let degrees = (radians * 180.0) / .pi\n \n var writerVideoInput: AVAssetWriterInput\n switch degrees {\n case 90:\n let rotateTransform = CGAffineTransform(rotationAngle: 0)\n writerVideoInput = AVAssetWriterInput.init(\n mediaType: AVMediaType.video,\n outputSettings: [\n AVVideoCodecKey: AVVideoCodecType.h264,\n AVVideoWidthKey: asset.tracks(withMediaType: AVMediaType.video)[0].naturalSize.height,\n AVVideoHeightKey: asset.tracks(withMediaType: AVMediaType.video)[0].naturalSize.width,\n AVVideoCompressionPropertiesKey: [\n AVVideoAverageBitRateKey: asset.tracks(withMediaType: AVMediaType.video)[0].estimatedDataRate,\n ],\n ]\n )\n writerVideoInput.expectsMediaDataInRealTime = false\n \n isFrameRotated = true\n writerVideoInput.transform = rotateTransform\n default:\n writerVideoInput = AVAssetWriterInput.init(\n mediaType: AVMediaType.video,\n outputSettings: [\n AVVideoCodecKey: AVVideoCodecType.h264,\n AVVideoWidthKey: asset.tracks(withMediaType: AVMediaType.video)[0].naturalSize.width,\n AVVideoHeightKey: asset.tracks(withMediaType: AVMediaType.video)[0].naturalSize.height,\n AVVideoCompressionPropertiesKey: [\n AVVideoAverageBitRateKey: asset.tracks(withMediaType: AVMediaType.video)[0].estimatedDataRate,\n ],\n ]\n )\n writerVideoInput.expectsMediaDataInRealTime = false\n isFrameRotated = false\n writerVideoInput.transform = asset.tracks(withMediaType: AVMediaType.video)[0].preferredTransform\n }\n let pixelBufferAdaptor = AVAssetWriterInputPixelBufferAdaptor(assetWriterInput: writerVideoInput, sourcePixelBufferAttributes: [kCVPixelBufferPixelFormatTypeKey as String: Int(kCVPixelFormatType_32BGRA)])\n \n writer.add(writerVideoInput)\n \n \n // prepare writer input for audio\n \n var writerAudioInput: AVAssetWriterInput! = nil\n if asset.tracks(withMediaType: AVMediaType.audio).count > 0 {\n let formatDesc: [Any] = asset.tracks(withMediaType: AVMediaType.audio)[0].formatDescriptions\n var channels: UInt32 = 1\n var sampleRate: Float64 = 44100.000000\n for i in 0 ..< formatDesc.count {\n guard let bobTheDesc: UnsafePointer = CMAudioFormatDescriptionGetStreamBasicDescription(formatDesc[i] as! CMAudioFormatDescription) else {\n continue\n }\n channels = bobTheDesc.pointee.mChannelsPerFrame\n sampleRate = bobTheDesc.pointee.mSampleRate\n break\n }\n writerAudioInput = AVAssetWriterInput.init(\n mediaType: AVMediaType.audio,\n outputSettings: [\n AVFormatIDKey: kAudioFormatMPEG4AAC,\n AVNumberOfChannelsKey: channels,\n AVSampleRateKey: sampleRate,\n AVEncoderBitRateKey: 128000,\n ]\n )\n writerAudioInput.expectsMediaDataInRealTime = true\n writer.add(writerAudioInput)\n }\n \n\n // write\n \n let videoQueue = DispatchQueue.init(label: \"videoQueue\")\n let audioQueue = DispatchQueue.init(label: \"audioQueue\")\n writer.startWriting()\n reader.startReading()\n writer.startSession(atSourceTime: CMTime.zero)\n\n // write video\n \n writerVideoInput.requestMediaDataWhenReady(on: videoQueue) {\n while writerVideoInput.isReadyForMoreMediaData {\n autoreleasepool {\n if let buffer = readerVideoOutput.copyNextSampleBuffer(),let pixelBuffer = CMSampleBufferGetImageBuffer(buffer) {\n frame += 1\n var ciImage = CIImage(cvPixelBuffer: pixelBuffer)\n if isFrameRotated {\n ciImage = ciImage.oriented(CGImagePropertyOrientation.right)\n }\n guard let outCIImage = processingFunction(ciImage) else { print(\"Video Processing Failed\") ; return }\n \n let presentationTime = CMSampleBufferGetOutputPresentationTimeStamp(buffer)\n var pixelBufferOut: CVPixelBuffer?\n CVPixelBufferPoolCreatePixelBuffer(kCFAllocatorDefault, pixelBufferAdaptor.pixelBufferPool!, &pixelBufferOut)\n self.ciContext.render(outCIImage, to: pixelBufferOut!)\n pixelBufferAdaptor.append(pixelBufferOut!, withPresentationTime: presentationTime)\n \n// if frame % 100 == 0 {\n// print(\"\\(frame) / \\(totalFrame) frames were processed..\")\n// }\n } else {\n writerVideoInput.markAsFinished()\n DispatchQueue.main.async {\n videoFinished = true\n writtingFinished()\n }\n }\n }\n }\n }\n if writerAudioInput != nil {\n writerAudioInput.requestMediaDataWhenReady(on: audioQueue) {\n while writerAudioInput.isReadyForMoreMediaData {\n autoreleasepool {\n let buffer = readerAudioOutput.copyNextSampleBuffer()\n if buffer != nil {\n writerAudioInput.append(buffer!)\n } else {\n writerAudioInput.markAsFinished()\n DispatchQueue.main.async {\n audioFinished = true\n writtingFinished()\n }\n }\n }\n }\n }\n }\n }\n \n func scaleMaskImage(maskCIImage:CIImage, originalCIImage:CIImage) -> CIImage {\n let scaledMaskCIImage = maskCIImage.resize(as: originalCIImage.extent.size)\n return scaledMaskCIImage\n }\n \n public func getCorrectOrientationUIImage(uiImage:UIImage) -> UIImage {\n var newImage = UIImage()\n switch uiImage.imageOrientation.rawValue {\n case 1:\n guard let orientedCIImage = CIImage(image: uiImage)?.oriented(CGImagePropertyOrientation.down),\n let cgImage = ciContext.createCGImage(orientedCIImage, from: orientedCIImage.extent) else { return uiImage}\n \n newImage = UIImage(cgImage: cgImage)\n case 3:\n guard let orientedCIImage = CIImage(image: uiImage)?.oriented(CGImagePropertyOrientation.right),\n let cgImage = ciContext.createCGImage(orientedCIImage, from: orientedCIImage.extent) else { return uiImage}\n newImage = UIImage(cgImage: cgImage)\n default:\n newImage = uiImage\n }\n return newImage\n }\n}\n \nextension CIImage {\n func resize(as size: CGSize) -> CIImage {\n let selfSize = extent.size\n let transform = CGAffineTransform(scaleX: size.width / selfSize.width, y: size.height / selfSize.height)\n return transformed(by: transform)\n }\n}\n\nextension CGPoint {\n func toVector(image: CIImage) -> CIVector {\n return CIVector(x: x, y: image.extent.height-y)\n }\n}\n\n" }, { "path": "Sources/SemanticImage/segmentation.mlmodelc/metadata.json", "content": "[\n {\n \"metadataOutputVersion\" : \"3.0\",\n \"storagePrecision\" : \"Float32\",\n \"outputSchema\" : [\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p0\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p1\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p2\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p3\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p4\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p5\"\n },\n {\n \"height\" : \"320\",\n \"colorspace\" : \"Grayscale\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"0\",\n \"formattedType\" : \"Image (Grayscale 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"out_p6\"\n }\n ],\n \"modelParameters\" : [\n\n ],\n \"specificationVersion\" : 5,\n \"computePrecision\" : \"Float16\",\n \"isUpdatable\" : \"0\",\n \"availability\" : {\n \"macOS\" : \"11.0\",\n \"tvOS\" : \"14.0\",\n \"watchOS\" : \"7.0\",\n \"iOS\" : \"14.0\",\n \"macCatalyst\" : \"14.0\"\n },\n \"neuralNetworkLayerTypeHistogram\" : {\n \"Concat\" : 51,\n \"ActivationLinear\" : 7,\n \"ActivationReLU\" : 112,\n \"UpsampleBiLinear\" : 38,\n \"Add\" : 11,\n \"BatchNorm\" : 112,\n \"Convolution\" : 119,\n \"ActivationSigmoid\" : 7,\n \"PoolingMax\" : 33\n },\n \"modelType\" : {\n \"name\" : \"MLModelType_neuralNetwork\"\n },\n \"userDefinedMetadata\" : {\n \"com.github.apple.coremltools.version\" : \"4.1\",\n \"com.github.apple.coremltools.source\" : \"torch==1.8.1+cu101\"\n },\n \"generatedClassName\" : \"segmentation\",\n \"inputSchema\" : [\n {\n \"height\" : \"320\",\n \"colorspace\" : \"RGB\",\n \"isOptional\" : \"0\",\n \"width\" : \"320\",\n \"isColor\" : \"1\",\n \"formattedType\" : \"Image (Color 320 Ɨ 320)\",\n \"hasSizeFlexibility\" : \"0\",\n \"type\" : \"Image\",\n \"shortDescription\" : \"\",\n \"name\" : \"in_0\"\n }\n ],\n \"method\" : \"predict\"\n }\n]" }, { "path": "Sources/SemanticImage/segmentation.mlmodelc/model.espresso.net", "content": "{\n \"layers\" : [\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.2\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.4\",\n \"blob_weights\" : 3,\n \"K\" : 3,\n \"blob_biases\" : 1,\n \"name\" : \"input.2\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"in_0\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.5\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.7\",\n \"blob_weights\" : 7,\n \"K\" : 64,\n \"blob_biases\" : 5,\n \"name\" : \"input.5\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.4\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.8\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.8\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.8\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.7\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.9\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.11\",\n \"blob_weights\" : 11,\n \"K\" : 16,\n \"blob_biases\" : 9,\n \"name\" : \"input.9\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.8\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.12\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.12\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.12\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.11\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.13\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.15\",\n \"blob_weights\" : 15,\n \"K\" : 16,\n \"blob_biases\" : 13,\n \"name\" : \"input.13\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.12\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.16\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.16\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.16\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.15\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.17\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.19\",\n \"blob_weights\" : 19,\n \"K\" : 16,\n \"blob_biases\" : 17,\n \"name\" : \"input.17\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.16\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.20\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.20\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.20\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.19\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.21\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.23\",\n \"blob_weights\" : 23,\n \"K\" : 16,\n \"blob_biases\" : 21,\n \"name\" : \"input.21\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.20\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.24\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.24\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.24\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.23\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.25\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.27\",\n \"blob_weights\" : 27,\n \"K\" : 16,\n \"blob_biases\" : 25,\n \"name\" : \"input.25\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.24\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.28\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx7.1\",\n \"blob_weights\" : 31,\n \"K\" : 16,\n \"blob_biases\" : 29,\n \"name\" : \"input.28\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.27\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.30\",\n \"top\" : \"input.30\",\n \"type\" : \"concat\",\n \"name\" : \"input.30\",\n \"bottom\" : \"hx7.1,input.27\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.31\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.33\",\n \"blob_weights\" : 35,\n \"K\" : 32,\n \"blob_biases\" : 33,\n \"name\" : \"input.31\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.30\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx6dup.1\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.33\",\n \"debug_info\" : \"hx6dup.1\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx6dup.1\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.34\",\n \"top\" : \"input.34\",\n \"type\" : \"concat\",\n \"name\" : \"input.34\",\n \"bottom\" : \"hx6dup.1,input.23\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.35\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.37\",\n \"blob_weights\" : 39,\n \"K\" : 32,\n \"blob_biases\" : 37,\n \"name\" : \"input.35\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.34\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx5dup.1\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.37\",\n \"debug_info\" : \"hx5dup.1\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx5dup.1\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.38\",\n \"top\" : \"input.38\",\n \"type\" : \"concat\",\n \"name\" : \"input.38\",\n \"bottom\" : \"hx5dup.1,input.19\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.39\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.41\",\n \"blob_weights\" : 43,\n \"K\" : 32,\n \"blob_biases\" : 41,\n \"name\" : \"input.39\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.38\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup.1\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.41\",\n \"debug_info\" : \"hx4dup.1\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.1\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.42\",\n \"top\" : \"input.42\",\n \"type\" : \"concat\",\n \"name\" : \"input.42\",\n \"bottom\" : \"hx4dup.1,input.15\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.43\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.45\",\n \"blob_weights\" : 47,\n \"K\" : 32,\n \"blob_biases\" : 45,\n \"name\" : \"input.43\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.42\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.1\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.45\",\n \"debug_info\" : \"hx3dup.1\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.1\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.46\",\n \"top\" : \"input.46\",\n \"type\" : \"concat\",\n \"name\" : \"input.46\",\n \"bottom\" : \"hx3dup.1,input.11\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.47\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.49\",\n \"blob_weights\" : 51,\n \"K\" : 32,\n \"blob_biases\" : 49,\n \"name\" : \"input.47\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.46\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.1\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.49\",\n \"debug_info\" : \"hx2dup.1\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.1\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.50\",\n \"top\" : \"input.50\",\n \"type\" : \"concat\",\n \"name\" : \"input.50\",\n \"bottom\" : \"hx2dup.1,input.7\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.51\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.1\",\n \"blob_weights\" : 55,\n \"K\" : 32,\n \"blob_biases\" : 53,\n \"name\" : \"input.51\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.50\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.4,hx1d.1\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.53\",\n \"top\" : \"input.53\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.53\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.54\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.54\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.54\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.53\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.55\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.57\",\n \"blob_weights\" : 59,\n \"K\" : 64,\n \"blob_biases\" : 57,\n \"name\" : \"input.55\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.54\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.58\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.60\",\n \"blob_weights\" : 63,\n \"K\" : 64,\n \"blob_biases\" : 61,\n \"name\" : \"input.58\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.57\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.61\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.61\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.61\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.60\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.62\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.64\",\n \"blob_weights\" : 67,\n \"K\" : 16,\n \"blob_biases\" : 65,\n \"name\" : \"input.62\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.61\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.65\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.65\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.65\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.64\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.66\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.68\",\n \"blob_weights\" : 71,\n \"K\" : 16,\n \"blob_biases\" : 69,\n \"name\" : \"input.66\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.65\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.69\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.69\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.69\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.68\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.70\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.72\",\n \"blob_weights\" : 75,\n \"K\" : 16,\n \"blob_biases\" : 73,\n \"name\" : \"input.70\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.69\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.73\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.73\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.73\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.72\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.74\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.76\",\n \"blob_weights\" : 79,\n \"K\" : 16,\n \"blob_biases\" : 77,\n \"name\" : \"input.74\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.73\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.77\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx6.1\",\n \"blob_weights\" : 83,\n \"K\" : 16,\n \"blob_biases\" : 81,\n \"name\" : \"input.77\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.76\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.79\",\n \"top\" : \"input.79\",\n \"type\" : \"concat\",\n \"name\" : \"input.79\",\n \"bottom\" : \"hx6.1,input.76\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.80\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.82\",\n \"blob_weights\" : 87,\n \"K\" : 32,\n \"blob_biases\" : 85,\n \"name\" : \"input.80\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.79\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx5dup.2\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.82\",\n \"debug_info\" : \"hx5dup.2\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx5dup.2\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.83\",\n \"top\" : \"input.83\",\n \"type\" : \"concat\",\n \"name\" : \"input.83\",\n \"bottom\" : \"hx5dup.2,input.72\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.84\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.86\",\n \"blob_weights\" : 91,\n \"K\" : 32,\n \"blob_biases\" : 89,\n \"name\" : \"input.84\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.83\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup.2\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.86\",\n \"debug_info\" : \"hx4dup.2\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.2\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.87\",\n \"top\" : \"input.87\",\n \"type\" : \"concat\",\n \"name\" : \"input.87\",\n \"bottom\" : \"hx4dup.2,input.68\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.88\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.90\",\n \"blob_weights\" : 95,\n \"K\" : 32,\n \"blob_biases\" : 93,\n \"name\" : \"input.88\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.87\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.2\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.90\",\n \"debug_info\" : \"hx3dup.2\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.2\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.91\",\n \"top\" : \"input.91\",\n \"type\" : \"concat\",\n \"name\" : \"input.91\",\n \"bottom\" : \"hx3dup.2,input.64\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.92\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.94\",\n \"blob_weights\" : 99,\n \"K\" : 32,\n \"blob_biases\" : 97,\n \"name\" : \"input.92\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.91\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.2\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.94\",\n \"debug_info\" : \"hx2dup.2\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.2\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.95\",\n \"top\" : \"input.95\",\n \"type\" : \"concat\",\n \"name\" : \"input.95\",\n \"bottom\" : \"hx2dup.2,input.60\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.96\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.2\",\n \"blob_weights\" : 103,\n \"K\" : 32,\n \"blob_biases\" : 101,\n \"name\" : \"input.96\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.95\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.57,hx1d.2\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.98\",\n \"top\" : \"input.98\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.98\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.99\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.99\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.99\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.98\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.100\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.102\",\n \"blob_weights\" : 107,\n \"K\" : 64,\n \"blob_biases\" : 105,\n \"name\" : \"input.100\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.99\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.103\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.105\",\n \"blob_weights\" : 111,\n \"K\" : 64,\n \"blob_biases\" : 109,\n \"name\" : \"input.103\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.102\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.106\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.106\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.106\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.105\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.107\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.109\",\n \"blob_weights\" : 115,\n \"K\" : 16,\n \"blob_biases\" : 113,\n \"name\" : \"input.107\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.106\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.110\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.110\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.110\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.109\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.111\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.113\",\n \"blob_weights\" : 119,\n \"K\" : 16,\n \"blob_biases\" : 117,\n \"name\" : \"input.111\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.110\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.114\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.114\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.114\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.113\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.115\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.117\",\n \"blob_weights\" : 123,\n \"K\" : 16,\n \"blob_biases\" : 121,\n \"name\" : \"input.115\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.114\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.118\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx5.1\",\n \"blob_weights\" : 127,\n \"K\" : 16,\n \"blob_biases\" : 125,\n \"name\" : \"input.118\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.117\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.120\",\n \"top\" : \"input.120\",\n \"type\" : \"concat\",\n \"name\" : \"input.120\",\n \"bottom\" : \"hx5.1,input.117\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.121\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.123\",\n \"blob_weights\" : 131,\n \"K\" : 32,\n \"blob_biases\" : 129,\n \"name\" : \"input.121\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.120\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup.3\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.123\",\n \"debug_info\" : \"hx4dup.3\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.3\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.124\",\n \"top\" : \"input.124\",\n \"type\" : \"concat\",\n \"name\" : \"input.124\",\n \"bottom\" : \"hx4dup.3,input.113\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.125\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.127\",\n \"blob_weights\" : 135,\n \"K\" : 32,\n \"blob_biases\" : 133,\n \"name\" : \"input.125\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.124\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.3\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.127\",\n \"debug_info\" : \"hx3dup.3\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.3\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.128\",\n \"top\" : \"input.128\",\n \"type\" : \"concat\",\n \"name\" : \"input.128\",\n \"bottom\" : \"hx3dup.3,input.109\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.129\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.131\",\n \"blob_weights\" : 139,\n \"K\" : 32,\n \"blob_biases\" : 137,\n \"name\" : \"input.129\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.128\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.3\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.131\",\n \"debug_info\" : \"hx2dup.3\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.3\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.132\",\n \"top\" : \"input.132\",\n \"type\" : \"concat\",\n \"name\" : \"input.132\",\n \"bottom\" : \"hx2dup.3,input.105\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.133\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.3\",\n \"blob_weights\" : 143,\n \"K\" : 32,\n \"blob_biases\" : 141,\n \"name\" : \"input.133\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.132\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.102,hx1d.3\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.135\",\n \"top\" : \"input.135\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.135\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.136\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.136\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.136\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.135\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.137\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.139\",\n \"blob_weights\" : 147,\n \"K\" : 64,\n \"blob_biases\" : 145,\n \"name\" : \"input.137\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.136\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.140\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.142\",\n \"blob_weights\" : 151,\n \"K\" : 64,\n \"blob_biases\" : 149,\n \"name\" : \"input.140\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.139\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.143\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.143\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.143\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.142\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.144\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.146\",\n \"blob_weights\" : 155,\n \"K\" : 16,\n \"blob_biases\" : 153,\n \"name\" : \"input.144\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.143\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.147\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.147\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.147\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.146\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.148\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.150\",\n \"blob_weights\" : 159,\n \"K\" : 16,\n \"blob_biases\" : 157,\n \"name\" : \"input.148\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.147\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.151\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx4.1\",\n \"blob_weights\" : 163,\n \"K\" : 16,\n \"blob_biases\" : 161,\n \"name\" : \"input.151\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.150\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.153\",\n \"top\" : \"input.153\",\n \"type\" : \"concat\",\n \"name\" : \"input.153\",\n \"bottom\" : \"hx4.1,input.150\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.154\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.156\",\n \"blob_weights\" : 167,\n \"K\" : 32,\n \"blob_biases\" : 165,\n \"name\" : \"input.154\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.153\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.4\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.156\",\n \"debug_info\" : \"hx3dup.4\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.4\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.157\",\n \"top\" : \"input.157\",\n \"type\" : \"concat\",\n \"name\" : \"input.157\",\n \"bottom\" : \"hx3dup.4,input.146\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.158\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.160\",\n \"blob_weights\" : 171,\n \"K\" : 32,\n \"blob_biases\" : 169,\n \"name\" : \"input.158\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.157\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.4\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.160\",\n \"debug_info\" : \"hx2dup.4\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.4\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.161\",\n \"top\" : \"input.161\",\n \"type\" : \"concat\",\n \"name\" : \"input.161\",\n \"bottom\" : \"hx2dup.4,input.142\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.162\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.4\",\n \"blob_weights\" : 175,\n \"K\" : 32,\n \"blob_biases\" : 173,\n \"name\" : \"input.162\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.161\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.139,hx1d.4\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.164\",\n \"top\" : \"input.164\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.164\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.165\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.165\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.165\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.164\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.166\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.168\",\n \"blob_weights\" : 179,\n \"K\" : 64,\n \"blob_biases\" : 177,\n \"name\" : \"input.166\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.165\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.169\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.171\",\n \"blob_weights\" : 183,\n \"K\" : 64,\n \"blob_biases\" : 181,\n \"name\" : \"input.169\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.168\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.172\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"input.174\",\n \"blob_weights\" : 187,\n \"K\" : 16,\n \"blob_biases\" : 185,\n \"name\" : \"input.172\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.171\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.175\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"input.177\",\n \"blob_weights\" : 191,\n \"K\" : 16,\n \"blob_biases\" : 189,\n \"name\" : \"input.175\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.174\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"pad_r\" : 8,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.178\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 8,\n \"pad_l\" : 8,\n \"top\" : \"hx4.2\",\n \"blob_weights\" : 195,\n \"K\" : 16,\n \"blob_biases\" : 193,\n \"name\" : \"input.178\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 8,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.177\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 8,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 8\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.180\",\n \"top\" : \"input.180\",\n \"type\" : \"concat\",\n \"name\" : \"input.180\",\n \"bottom\" : \"hx4.2,input.177\"\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.181\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"hx3d.1\",\n \"blob_weights\" : 199,\n \"K\" : 32,\n \"blob_biases\" : 197,\n \"name\" : \"input.181\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.180\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.183\",\n \"top\" : \"input.183\",\n \"type\" : \"concat\",\n \"name\" : \"input.183\",\n \"bottom\" : \"hx3d.1,input.174\"\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.184\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx2d.1\",\n \"blob_weights\" : 203,\n \"K\" : 32,\n \"blob_biases\" : 201,\n \"name\" : \"input.184\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.183\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.186\",\n \"top\" : \"input.186\",\n \"type\" : \"concat\",\n \"name\" : \"input.186\",\n \"bottom\" : \"hx2d.1,input.171\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.187\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.5\",\n \"blob_weights\" : 207,\n \"K\" : 32,\n \"blob_biases\" : 205,\n \"name\" : \"input.187\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.186\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.168,hx1d.5\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.189\",\n \"top\" : \"input.189\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.189\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.190\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.190\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.190\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.189\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.191\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.193\",\n \"blob_weights\" : 211,\n \"K\" : 64,\n \"blob_biases\" : 209,\n \"name\" : \"input.191\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.190\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.194\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.196\",\n \"blob_weights\" : 215,\n \"K\" : 64,\n \"blob_biases\" : 213,\n \"name\" : \"input.194\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.193\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.197\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"input.199\",\n \"blob_weights\" : 219,\n \"K\" : 16,\n \"blob_biases\" : 217,\n \"name\" : \"input.197\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.196\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.200\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"input.202\",\n \"blob_weights\" : 223,\n \"K\" : 16,\n \"blob_biases\" : 221,\n \"name\" : \"input.200\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.199\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"pad_r\" : 8,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.203\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 8,\n \"pad_l\" : 8,\n \"top\" : \"hx4.3\",\n \"blob_weights\" : 227,\n \"K\" : 16,\n \"blob_biases\" : 225,\n \"name\" : \"input.203\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 8,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.202\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 8,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 8\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.205\",\n \"top\" : \"input.205\",\n \"type\" : \"concat\",\n \"name\" : \"input.205\",\n \"bottom\" : \"hx4.3,input.202\"\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.206\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"hx3d.2\",\n \"blob_weights\" : 231,\n \"K\" : 32,\n \"blob_biases\" : 229,\n \"name\" : \"input.206\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.205\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.208\",\n \"top\" : \"input.208\",\n \"type\" : \"concat\",\n \"name\" : \"input.208\",\n \"bottom\" : \"hx3d.2,input.199\"\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.209\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx2d.2\",\n \"blob_weights\" : 235,\n \"K\" : 32,\n \"blob_biases\" : 233,\n \"name\" : \"input.209\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.208\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.211\",\n \"top\" : \"input.211\",\n \"type\" : \"concat\",\n \"name\" : \"input.211\",\n \"bottom\" : \"hx2d.2,input.196\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.212\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.6\",\n \"blob_weights\" : 239,\n \"K\" : 32,\n \"blob_biases\" : 237,\n \"name\" : \"input.212\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.211\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.193,hx1d.6\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.214\",\n \"top\" : \"input.214\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.214\",\n \"beta\" : 0\n },\n {\n \"top\" : \"hx6up\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.214\",\n \"debug_info\" : \"hx6up\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx6up\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.215\",\n \"top\" : \"input.215\",\n \"type\" : \"concat\",\n \"name\" : \"input.215\",\n \"bottom\" : \"hx6up,input.189\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.216\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.218\",\n \"blob_weights\" : 243,\n \"K\" : 128,\n \"blob_biases\" : 241,\n \"name\" : \"input.216\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.215\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.219\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.221\",\n \"blob_weights\" : 247,\n \"K\" : 64,\n \"blob_biases\" : 245,\n \"name\" : \"input.219\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.218\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.222\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"input.224\",\n \"blob_weights\" : 251,\n \"K\" : 16,\n \"blob_biases\" : 249,\n \"name\" : \"input.222\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.221\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.225\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"input.227\",\n \"blob_weights\" : 255,\n \"K\" : 16,\n \"blob_biases\" : 253,\n \"name\" : \"input.225\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.224\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"pad_r\" : 8,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.228\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 8,\n \"pad_l\" : 8,\n \"top\" : \"hx4.4\",\n \"blob_weights\" : 259,\n \"K\" : 16,\n \"blob_biases\" : 257,\n \"name\" : \"input.228\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 8,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.227\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 8,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 8\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.230\",\n \"top\" : \"input.230\",\n \"type\" : \"concat\",\n \"name\" : \"input.230\",\n \"bottom\" : \"hx4.4,input.227\"\n },\n {\n \"pad_r\" : 4,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.231\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 4,\n \"pad_l\" : 4,\n \"top\" : \"hx3d\",\n \"blob_weights\" : 263,\n \"K\" : 32,\n \"blob_biases\" : 261,\n \"name\" : \"input.231\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 4,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.230\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 4,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 4\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.233\",\n \"top\" : \"input.233\",\n \"type\" : \"concat\",\n \"name\" : \"input.233\",\n \"bottom\" : \"hx3d,input.224\"\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.234\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx2d\",\n \"blob_weights\" : 267,\n \"K\" : 32,\n \"blob_biases\" : 265,\n \"name\" : \"input.234\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.233\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.236\",\n \"top\" : \"input.236\",\n \"type\" : \"concat\",\n \"name\" : \"input.236\",\n \"bottom\" : \"hx2d,input.221\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.237\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.7\",\n \"blob_weights\" : 271,\n \"K\" : 32,\n \"blob_biases\" : 269,\n \"name\" : \"input.237\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.236\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.218,hx1d.7\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.239\",\n \"top\" : \"input.239\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.239\",\n \"beta\" : 0\n },\n {\n \"top\" : \"hx5dup.3\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.239\",\n \"debug_info\" : \"hx5dup.3\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx5dup.3\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.240\",\n \"top\" : \"input.240\",\n \"type\" : \"concat\",\n \"name\" : \"input.240\",\n \"bottom\" : \"hx5dup.3,input.164\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.241\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.243\",\n \"blob_weights\" : 275,\n \"K\" : 128,\n \"blob_biases\" : 273,\n \"name\" : \"input.241\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.240\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.244\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.246\",\n \"blob_weights\" : 279,\n \"K\" : 64,\n \"blob_biases\" : 277,\n \"name\" : \"input.244\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.243\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.247\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.247\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.247\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.246\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.248\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.250\",\n \"blob_weights\" : 283,\n \"K\" : 16,\n \"blob_biases\" : 281,\n \"name\" : \"input.248\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.247\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.251\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.251\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.251\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.250\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.252\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.254\",\n \"blob_weights\" : 287,\n \"K\" : 16,\n \"blob_biases\" : 285,\n \"name\" : \"input.252\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.251\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.255\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx4\",\n \"blob_weights\" : 291,\n \"K\" : 16,\n \"blob_biases\" : 289,\n \"name\" : \"input.255\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.254\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.257\",\n \"top\" : \"input.257\",\n \"type\" : \"concat\",\n \"name\" : \"input.257\",\n \"bottom\" : \"hx4,input.254\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.258\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.260\",\n \"blob_weights\" : 295,\n \"K\" : 32,\n \"blob_biases\" : 293,\n \"name\" : \"input.258\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.257\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.5\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.260\",\n \"debug_info\" : \"hx3dup.5\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.5\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.261\",\n \"top\" : \"input.261\",\n \"type\" : \"concat\",\n \"name\" : \"input.261\",\n \"bottom\" : \"hx3dup.5,input.250\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.262\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.264\",\n \"blob_weights\" : 299,\n \"K\" : 32,\n \"blob_biases\" : 297,\n \"name\" : \"input.262\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.261\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.5\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.264\",\n \"debug_info\" : \"hx2dup.5\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.5\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.265\",\n \"top\" : \"input.265\",\n \"type\" : \"concat\",\n \"name\" : \"input.265\",\n \"bottom\" : \"hx2dup.5,input.246\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.266\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.8\",\n \"blob_weights\" : 303,\n \"K\" : 32,\n \"blob_biases\" : 301,\n \"name\" : \"input.266\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.265\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.243,hx1d.8\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.268\",\n \"top\" : \"input.268\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.268\",\n \"beta\" : 0\n },\n {\n \"top\" : \"hx4dup.4\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.268\",\n \"debug_info\" : \"hx4dup.4\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.4\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.269\",\n \"top\" : \"input.269\",\n \"type\" : \"concat\",\n \"name\" : \"input.269\",\n \"bottom\" : \"hx4dup.4,input.135\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.270\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.272\",\n \"blob_weights\" : 307,\n \"K\" : 128,\n \"blob_biases\" : 305,\n \"name\" : \"input.270\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.269\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.273\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.275\",\n \"blob_weights\" : 311,\n \"K\" : 64,\n \"blob_biases\" : 309,\n \"name\" : \"input.273\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.272\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.276\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.276\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.276\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.275\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.277\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.279\",\n \"blob_weights\" : 315,\n \"K\" : 16,\n \"blob_biases\" : 313,\n \"name\" : \"input.277\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.276\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.280\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.280\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.280\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.279\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.281\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.283\",\n \"blob_weights\" : 319,\n \"K\" : 16,\n \"blob_biases\" : 317,\n \"name\" : \"input.281\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.280\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.284\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.284\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.284\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.283\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.285\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.287\",\n \"blob_weights\" : 323,\n \"K\" : 16,\n \"blob_biases\" : 321,\n \"name\" : \"input.285\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.284\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.288\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx5\",\n \"blob_weights\" : 327,\n \"K\" : 16,\n \"blob_biases\" : 325,\n \"name\" : \"input.288\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.287\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.290\",\n \"top\" : \"input.290\",\n \"type\" : \"concat\",\n \"name\" : \"input.290\",\n \"bottom\" : \"hx5,input.287\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.291\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.293\",\n \"blob_weights\" : 331,\n \"K\" : 32,\n \"blob_biases\" : 329,\n \"name\" : \"input.291\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.290\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup.5\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.293\",\n \"debug_info\" : \"hx4dup.5\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.5\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.294\",\n \"top\" : \"input.294\",\n \"type\" : \"concat\",\n \"name\" : \"input.294\",\n \"bottom\" : \"hx4dup.5,input.283\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.295\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.297\",\n \"blob_weights\" : 335,\n \"K\" : 32,\n \"blob_biases\" : 333,\n \"name\" : \"input.295\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.294\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.6\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.297\",\n \"debug_info\" : \"hx3dup.6\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.6\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.298\",\n \"top\" : \"input.298\",\n \"type\" : \"concat\",\n \"name\" : \"input.298\",\n \"bottom\" : \"hx3dup.6,input.279\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.299\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.301\",\n \"blob_weights\" : 339,\n \"K\" : 32,\n \"blob_biases\" : 337,\n \"name\" : \"input.299\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.298\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.6\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.301\",\n \"debug_info\" : \"hx2dup.6\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.6\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.302\",\n \"top\" : \"input.302\",\n \"type\" : \"concat\",\n \"name\" : \"input.302\",\n \"bottom\" : \"hx2dup.6,input.275\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.303\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.9\",\n \"blob_weights\" : 343,\n \"K\" : 32,\n \"blob_biases\" : 341,\n \"name\" : \"input.303\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.302\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.272,hx1d.9\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.305\",\n \"top\" : \"input.305\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.305\",\n \"beta\" : 0\n },\n {\n \"top\" : \"hx3dup.7\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.305\",\n \"debug_info\" : \"hx3dup.7\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.7\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.306\",\n \"top\" : \"input.306\",\n \"type\" : \"concat\",\n \"name\" : \"input.306\",\n \"bottom\" : \"hx3dup.7,input.98\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.307\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.309\",\n \"blob_weights\" : 347,\n \"K\" : 128,\n \"blob_biases\" : 345,\n \"name\" : \"input.307\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.306\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.310\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.312\",\n \"blob_weights\" : 351,\n \"K\" : 64,\n \"blob_biases\" : 349,\n \"name\" : \"input.310\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.309\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.313\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.313\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.313\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.312\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.314\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.316\",\n \"blob_weights\" : 355,\n \"K\" : 16,\n \"blob_biases\" : 353,\n \"name\" : \"input.314\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.313\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.317\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.317\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.317\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.316\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.318\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.320\",\n \"blob_weights\" : 359,\n \"K\" : 16,\n \"blob_biases\" : 357,\n \"name\" : \"input.318\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.317\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.321\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.321\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.321\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.320\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.322\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.324\",\n \"blob_weights\" : 363,\n \"K\" : 16,\n \"blob_biases\" : 361,\n \"name\" : \"input.322\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.321\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.325\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.325\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.325\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.324\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.326\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.328\",\n \"blob_weights\" : 367,\n \"K\" : 16,\n \"blob_biases\" : 365,\n \"name\" : \"input.326\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.325\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.329\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx6\",\n \"blob_weights\" : 371,\n \"K\" : 16,\n \"blob_biases\" : 369,\n \"name\" : \"input.329\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.328\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.331\",\n \"top\" : \"input.331\",\n \"type\" : \"concat\",\n \"name\" : \"input.331\",\n \"bottom\" : \"hx6,input.328\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.332\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.334\",\n \"blob_weights\" : 375,\n \"K\" : 32,\n \"blob_biases\" : 373,\n \"name\" : \"input.332\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.331\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx5dup.4\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.334\",\n \"debug_info\" : \"hx5dup.4\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx5dup.4\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.335\",\n \"top\" : \"input.335\",\n \"type\" : \"concat\",\n \"name\" : \"input.335\",\n \"bottom\" : \"hx5dup.4,input.324\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.336\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.338\",\n \"blob_weights\" : 379,\n \"K\" : 32,\n \"blob_biases\" : 377,\n \"name\" : \"input.336\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.335\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup.6\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.338\",\n \"debug_info\" : \"hx4dup.6\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup.6\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.339\",\n \"top\" : \"input.339\",\n \"type\" : \"concat\",\n \"name\" : \"input.339\",\n \"bottom\" : \"hx4dup.6,input.320\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.340\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.342\",\n \"blob_weights\" : 383,\n \"K\" : 32,\n \"blob_biases\" : 381,\n \"name\" : \"input.340\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.339\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup.8\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.342\",\n \"debug_info\" : \"hx3dup.8\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup.8\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.343\",\n \"top\" : \"input.343\",\n \"type\" : \"concat\",\n \"name\" : \"input.343\",\n \"bottom\" : \"hx3dup.8,input.316\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.344\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.346\",\n \"blob_weights\" : 387,\n \"K\" : 32,\n \"blob_biases\" : 385,\n \"name\" : \"input.344\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.343\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup.7\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.346\",\n \"debug_info\" : \"hx2dup.7\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.7\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.347\",\n \"top\" : \"input.347\",\n \"type\" : \"concat\",\n \"name\" : \"input.347\",\n \"bottom\" : \"hx2dup.7,input.312\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.348\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d.10\",\n \"blob_weights\" : 391,\n \"K\" : 32,\n \"blob_biases\" : 389,\n \"name\" : \"input.348\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.347\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.309,hx1d.10\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.350\",\n \"top\" : \"input.350\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.350\",\n \"beta\" : 0\n },\n {\n \"top\" : \"hx2dup.8\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.350\",\n \"debug_info\" : \"hx2dup.8\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup.8\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.351\",\n \"top\" : \"input.351\",\n \"type\" : \"concat\",\n \"name\" : \"input.351\",\n \"bottom\" : \"hx2dup.8,input.53\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.352\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.354\",\n \"blob_weights\" : 395,\n \"K\" : 128,\n \"blob_biases\" : 393,\n \"name\" : \"input.352\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.351\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.355\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.357\",\n \"blob_weights\" : 399,\n \"K\" : 64,\n \"blob_biases\" : 397,\n \"name\" : \"input.355\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.354\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.358\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.358\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.358\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.357\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.359\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.361\",\n \"blob_weights\" : 403,\n \"K\" : 16,\n \"blob_biases\" : 401,\n \"name\" : \"input.359\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.358\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.362\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.362\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.362\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.361\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.363\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.365\",\n \"blob_weights\" : 407,\n \"K\" : 16,\n \"blob_biases\" : 405,\n \"name\" : \"input.363\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.362\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.366\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.366\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.366\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.365\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.367\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.369\",\n \"blob_weights\" : 411,\n \"K\" : 16,\n \"blob_biases\" : 409,\n \"name\" : \"input.367\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.366\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.370\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.370\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.370\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.369\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.371\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.373\",\n \"blob_weights\" : 415,\n \"K\" : 16,\n \"blob_biases\" : 413,\n \"name\" : \"input.371\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.370\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 0,\n \"debug_info\" : \"input.374\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"size_x\" : 2,\n \"top\" : \"input.374\",\n \"top_shape_style\" : 0,\n \"stride_x\" : 2,\n \"avg_or_max\" : 1,\n \"average_count_exclude_padding\" : 1,\n \"type\" : \"pool\",\n \"name\" : \"input.374\",\n \"pad_t\" : 0,\n \"stride_y\" : 2,\n \"bottom\" : \"input.373\",\n \"weights\" : {\n\n },\n \"pad_mode\" : 2,\n \"size_y\" : 2,\n \"pad_value\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.375\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.377\",\n \"blob_weights\" : 419,\n \"K\" : 16,\n \"blob_biases\" : 417,\n \"name\" : \"input.375\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.374\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 2,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.378\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 2,\n \"pad_l\" : 2,\n \"top\" : \"hx7\",\n \"blob_weights\" : 423,\n \"K\" : 16,\n \"blob_biases\" : 421,\n \"name\" : \"input.378\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 2,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.377\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"dilation_x\" : 2,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1,\n \"dilation_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.380\",\n \"top\" : \"input.380\",\n \"type\" : \"concat\",\n \"name\" : \"input.380\",\n \"bottom\" : \"hx7,input.377\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.381\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.383\",\n \"blob_weights\" : 427,\n \"K\" : 32,\n \"blob_biases\" : 425,\n \"name\" : \"input.381\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.380\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx6dup\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.383\",\n \"debug_info\" : \"hx6dup\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx6dup\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.384\",\n \"top\" : \"input.384\",\n \"type\" : \"concat\",\n \"name\" : \"input.384\",\n \"bottom\" : \"hx6dup,input.373\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.385\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.387\",\n \"blob_weights\" : 431,\n \"K\" : 32,\n \"blob_biases\" : 429,\n \"name\" : \"input.385\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.384\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx5dup\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.387\",\n \"debug_info\" : \"hx5dup\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx5dup\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.388\",\n \"top\" : \"input.388\",\n \"type\" : \"concat\",\n \"name\" : \"input.388\",\n \"bottom\" : \"hx5dup,input.369\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.389\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.391\",\n \"blob_weights\" : 435,\n \"K\" : 32,\n \"blob_biases\" : 433,\n \"name\" : \"input.389\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.388\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx4dup\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.391\",\n \"debug_info\" : \"hx4dup\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx4dup\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.392\",\n \"top\" : \"input.392\",\n \"type\" : \"concat\",\n \"name\" : \"input.392\",\n \"bottom\" : \"hx4dup,input.365\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.393\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.395\",\n \"blob_weights\" : 439,\n \"K\" : 32,\n \"blob_biases\" : 437,\n \"name\" : \"input.393\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.392\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx3dup\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.395\",\n \"debug_info\" : \"hx3dup\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx3dup\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.396\",\n \"top\" : \"input.396\",\n \"type\" : \"concat\",\n \"name\" : \"input.396\",\n \"bottom\" : \"hx3dup,input.361\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.397\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.399\",\n \"blob_weights\" : 443,\n \"K\" : 32,\n \"blob_biases\" : 441,\n \"name\" : \"input.397\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 16,\n \"bottom\" : \"input.396\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"hx2dup\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.399\",\n \"debug_info\" : \"hx2dup\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"hx2dup\",\n \"scaling_factor_y\" : 2\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.400\",\n \"top\" : \"input.400\",\n \"type\" : \"concat\",\n \"name\" : \"input.400\",\n \"bottom\" : \"hx2dup,input.357\"\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 1,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.401\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"hx1d\",\n \"blob_weights\" : 447,\n \"K\" : 32,\n \"blob_biases\" : 445,\n \"name\" : \"input.401\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 64,\n \"bottom\" : \"input.400\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input.354,hx1d\",\n \"alpha\" : 1,\n \"operation\" : 0,\n \"weights\" : {\n\n },\n \"fused_relu\" : 0,\n \"debug_info\" : \"input.403\",\n \"top\" : \"input.403\",\n \"type\" : \"elementwise\",\n \"name\" : \"input.403\",\n \"beta\" : 0\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"tar\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"tar\",\n \"blob_weights\" : 451,\n \"K\" : 64,\n \"blob_biases\" : 449,\n \"name\" : \"tar\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.403\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.404\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.404\",\n \"blob_weights\" : 455,\n \"K\" : 64,\n \"blob_biases\" : 453,\n \"name\" : \"input.404\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.350\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"d2\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 2,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.404\",\n \"debug_info\" : \"d2\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"d2\",\n \"scaling_factor_y\" : 2\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.405\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.405\",\n \"blob_weights\" : 459,\n \"K\" : 64,\n \"blob_biases\" : 457,\n \"name\" : \"input.405\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.305\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"d3\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 4,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.405\",\n \"debug_info\" : \"d3\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"d3\",\n \"scaling_factor_y\" : 4\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.406\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.406\",\n \"blob_weights\" : 463,\n \"K\" : 64,\n \"blob_biases\" : 461,\n \"name\" : \"input.406\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.268\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"d4\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 8,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.406\",\n \"debug_info\" : \"d4\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"d4\",\n \"scaling_factor_y\" : 8\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.407\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.407\",\n \"blob_weights\" : 467,\n \"K\" : 64,\n \"blob_biases\" : 465,\n \"name\" : \"input.407\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.239\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"d5\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 16,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.407\",\n \"debug_info\" : \"d5\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"d5\",\n \"scaling_factor_y\" : 16\n },\n {\n \"pad_r\" : 1,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input.408\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 1,\n \"pad_l\" : 1,\n \"top\" : \"input.408\",\n \"blob_weights\" : 471,\n \"K\" : 64,\n \"blob_biases\" : 469,\n \"name\" : \"input.408\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 1,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.214\",\n \"weights\" : {\n\n },\n \"Nx\" : 3,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 3,\n \"n_parallel\" : 1\n },\n {\n \"top\" : \"d6\",\n \"use_fractional_scale_factors\" : false,\n \"weights\" : {\n\n },\n \"fractional_scaling_factor_y\" : 1,\n \"type\" : \"upsample\",\n \"scaling_factor_x\" : 32,\n \"mode\" : 1,\n \"align_corners\" : 0,\n \"bottom\" : \"input.408\",\n \"debug_info\" : \"d6\",\n \"fractional_scaling_factor_x\" : 1,\n \"is_legacy_mode\" : 0,\n \"name\" : \"d6\",\n \"scaling_factor_y\" : 32\n },\n {\n \"weights\" : {\n\n },\n \"debug_info\" : \"input.409\",\n \"top\" : \"input.409\",\n \"type\" : \"concat\",\n \"name\" : \"input.409\",\n \"bottom\" : \"tar,d2,d3,d4,d5,d6\"\n },\n {\n \"pad_r\" : 0,\n \"fused_relu\" : 0,\n \"fused_tanh\" : 0,\n \"debug_info\" : \"input\",\n \"pad_fill_mode\" : 0,\n \"pad_b\" : 0,\n \"pad_l\" : 0,\n \"top\" : \"input\",\n \"blob_weights\" : 475,\n \"K\" : 6,\n \"blob_biases\" : 473,\n \"name\" : \"input\",\n \"has_batch_norm\" : 0,\n \"type\" : \"convolution\",\n \"n_groups\" : 1,\n \"pad_t\" : 0,\n \"has_biases\" : 1,\n \"C\" : 1,\n \"bottom\" : \"input.409\",\n \"weights\" : {\n\n },\n \"Nx\" : 1,\n \"pad_mode\" : 0,\n \"pad_value\" : 0,\n \"Ny\" : 1,\n \"n_parallel\" : 1\n },\n {\n \"bottom\" : \"input\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2169\",\n \"top\" : \"out_a0\",\n \"type\" : \"activation\",\n \"name\" : \"2169\"\n },\n {\n \"bottom\" : \"tar\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2170\",\n \"top\" : \"out_a1\",\n \"type\" : \"activation\",\n \"name\" : \"2170\"\n },\n {\n \"bottom\" : \"d2\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2171\",\n \"top\" : \"out_a2\",\n \"type\" : \"activation\",\n \"name\" : \"2171\"\n },\n {\n \"bottom\" : \"d3\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2172\",\n \"top\" : \"out_a3\",\n \"type\" : \"activation\",\n \"name\" : \"2172\"\n },\n {\n \"bottom\" : \"d4\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2173\",\n \"top\" : \"out_a4\",\n \"type\" : \"activation\",\n \"name\" : \"2173\"\n },\n {\n \"bottom\" : \"d5\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2174\",\n \"top\" : \"out_a5\",\n \"type\" : \"activation\",\n \"name\" : \"2174\"\n },\n {\n \"bottom\" : \"d6\",\n \"weights\" : {\n\n },\n \"mode\" : 3,\n \"debug_info\" : \"2175\",\n \"top\" : \"out_a6\",\n \"type\" : \"activation\",\n \"name\" : \"2175\"\n },\n {\n \"bottom\" : \"out_a0\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p0\",\n \"top\" : \"out_p0\",\n \"type\" : \"activation\",\n \"name\" : \"out_p0\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a1\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p1\",\n \"top\" : \"out_p1\",\n \"type\" : \"activation\",\n \"name\" : \"out_p1\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a2\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p2\",\n \"top\" : \"out_p2\",\n \"type\" : \"activation\",\n \"name\" : \"out_p2\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a3\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p3\",\n \"top\" : \"out_p3\",\n \"type\" : \"activation\",\n \"name\" : \"out_p3\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a4\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p4\",\n \"top\" : \"out_p4\",\n \"type\" : \"activation\",\n \"name\" : \"out_p4\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a5\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p5\",\n \"top\" : \"out_p5\",\n \"type\" : \"activation\",\n \"name\" : \"out_p5\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n },\n {\n \"bottom\" : \"out_a6\",\n \"alpha\" : 255,\n \"beta\" : 0,\n \"mode\" : 6,\n \"weights\" : {\n\n },\n \"debug_info\" : \"out_p6\",\n \"top\" : \"out_p6\",\n \"type\" : \"activation\",\n \"name\" : \"out_p6\",\n \"attributes\" : {\n \"is_output\" : 1\n }\n }\n ],\n \"transform_params\" : {\n \"in_0\" : {\n \"bias_a\" : 0,\n \"bias_g\" : 0,\n \"bias_r\" : 0,\n \"bias_b\" : 0,\n \"center_mean\" : 0,\n \"is_network_bgr\" : 0,\n \"scale\" : 0.0039215688593685627\n }\n },\n \"storage\" : \"model.espresso.weights\",\n \"metadata_in_weights\" : [\n\n ],\n \"properties\" : {\n\n },\n \"analyses\" : {\n\n },\n \"format_version\" : 200\n}" }, { "path": "Sources/SemanticImage/segmentation.mlmodelc/model.espresso.shape", "content": "{\n \"layer_shapes\" : {\n \"input.95\" : {\n \"k\" : 32,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.354\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.41\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.369\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.383\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.403\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"tar\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.16\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx4.4\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.208\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx5dup.4\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.251\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.46\" : {\n \"k\" : 32,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.280\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx6dup.1\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_p6\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.76\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx6dup\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.373\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.105\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.388\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.8\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.408\" : {\n \"k\" : 1,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx2dup.1\" : {\n \"k\" : 16,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.227\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx5.1\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.27\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx1d.2\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.82\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.305\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx2dup.6\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.57\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.334\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx1d.10\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.392\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.124\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.139\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.87\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx3d\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.33\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx3dup.3\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.168\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.153\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.202\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.246\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.260\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.275\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx3dup.8\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.38\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.324\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.339\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.114\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx4dup.5\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.68\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.143\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx1d.3\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx2d\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.221\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx5dup.2\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.98\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.236\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.250\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.265\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.294\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.19\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.343\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.20\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.358\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.387\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.4\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.49\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.407\" : {\n \"k\" : 1,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.50\" : {\n \"k\" : 32,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.177\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.211\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.79\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.240\" : {\n \"k\" : 128,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx4dup\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx1d\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.284\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx2dup.4\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx1d.4\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.362\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.377\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.391\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.123\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.109\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx3dup.1\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"hx6.1\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"out_a0\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.196\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.230\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.61\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx3dup.6\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.309\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.91\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.338\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.113\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx4dup.3\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx2d.1\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.396\" : {\n \"k\" : 32,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.142\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.128\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.12\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.157\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.171\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.186\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx1d.5\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.42\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.264\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.279\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.293\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.313\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.72\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.328\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.342\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"out_a1\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.357\" : {\n \"k\" : 16,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"out_p0\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.132\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.406\" : {\n \"k\" : 1,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.147\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.161\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.190\" : {\n \"k\" : 64,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.23\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.254\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.269\" : {\n \"k\" : 128,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.283\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.298\" : {\n \"k\" : 32,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx2d.2\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx2dup.2\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.53\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.347\" : {\n \"k\" : 32,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.361\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.83\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx1d.6\" : {\n \"k\" : 64,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx2dup\" : {\n \"k\" : 16,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.180\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx2dup.7\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.215\" : {\n \"k\" : 128,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_a2\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"out_p1\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"hx3dup.4\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.34\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx7.1\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.351\" : {\n \"k\" : 128,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"in_0\" : {\n \"k\" : 3,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.366\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.380\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx4dup.1\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.64\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.127\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.400\" : {\n \"k\" : 32,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.395\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.156\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.205\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.94\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx4dup.6\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.69\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.15\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.312\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"hx1d.7\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.370\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.99\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.102\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.45\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.117\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.131\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx5dup.3\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.405\" : {\n \"k\" : 1,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.146\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.160\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_a3\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"out_p2\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.224\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.239\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.268\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.302\" : {\n \"k\" : 32,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.297\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx5dup\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.317\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.331\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.346\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx4\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.136\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.150\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.165\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx5\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx2dup.5\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.214\" : {\n \"k\" : 64,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx1d.8\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx6\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.243\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.272\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx3dup.2\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.86\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.287\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx7\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.321\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_a4\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.350\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"out_p3\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.365\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"hx3dup.7\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"hx3d.1\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.37\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"d2\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.199\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.233\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"d3\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"hx4dup.4\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"d4\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"d5\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"d6\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"hx5dup.1\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.384\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx1d.9\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.399\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.404\" : {\n \"k\" : 1,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.174\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.73\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.189\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_a5\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"out_p4\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.301\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.316\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.7\" : {\n \"k\" : 16,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"hx4.1\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx3d.2\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.24\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.374\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.106\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.120\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.135\" : {\n \"k\" : 64,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.409\" : {\n \"k\" : 6,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.164\" : {\n \"k\" : 64,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx2dup.3\" : {\n \"k\" : 16,\n \"w\" : 80,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 80\n },\n \"input.54\" : {\n \"k\" : 64,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"hx6up\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.193\" : {\n \"k\" : 64,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx3dup\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.257\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.30\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx4.2\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.306\" : {\n \"k\" : 128,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.320\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"hx2dup.8\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.335\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.60\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.110\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_a6\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"hx3dup.5\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"out_p5\" : {\n \"k\" : 1,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n },\n \"input.183\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.90\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.218\" : {\n \"k\" : 64,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"hx4.3\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"input.261\" : {\n \"k\" : 32,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.65\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.247\" : {\n \"k\" : 16,\n \"w\" : 20,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 20\n },\n \"input.11\" : {\n \"k\" : 16,\n \"w\" : 160,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 160\n },\n \"input.290\" : {\n \"k\" : 32,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx4dup.2\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.276\" : {\n \"k\" : 16,\n \"w\" : 40,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 40\n },\n \"input.325\" : {\n \"k\" : 16,\n \"w\" : 10,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 10\n },\n \"hx1d.1\" : {\n \"k\" : 64,\n \"w\" : 320,\n \"n\" : 1,\n \"_rank\" : 4,\n \"h\" : 320\n }\n }\n}" }, { "path": "Sources/SemanticImage/segmentation.swift", "content": "//\n// segmentation.swift\n//\n// This file was automatically generated and should not be edited.\n//\n\nimport CoreML\n\n\n/// Model Prediction Input Type\n@available(macOS 11.0, iOS 14.0, tvOS 14.0, watchOS 7.0, *)\nclass segmentationInput : MLFeatureProvider {\n\n /// in_0 as color (kCVPixelFormatType_32BGRA) image buffer, 320 pixels wide by 320 pixels high\n var in_0: CVPixelBuffer\n\n var featureNames: Set {\n get {\n return [\"in_0\"]\n }\n }\n \n func featureValue(for featureName: String) -> MLFeatureValue? {\n if (featureName == \"in_0\") {\n return MLFeatureValue(pixelBuffer: in_0)\n }\n return nil\n }\n \n init(in_0: CVPixelBuffer) {\n self.in_0 = in_0\n }\n\n convenience init(in_0With in_0: CGImage) throws {\n self.init(in_0: try MLFeatureValue(cgImage: in_0, pixelsWide: 320, pixelsHigh: 320, pixelFormatType: kCVPixelFormatType_32ARGB, options: nil).imageBufferValue!)\n }\n\n convenience init(in_0At in_0: URL) throws {\n self.init(in_0: try MLFeatureValue(imageAt: in_0, pixelsWide: 320, pixelsHigh: 320, pixelFormatType: kCVPixelFormatType_32ARGB, options: nil).imageBufferValue!)\n }\n\n func setIn_0(with in_0: CGImage) throws {\n self.in_0 = try MLFeatureValue(cgImage: in_0, pixelsWide: 320, pixelsHigh: 320, pixelFormatType: kCVPixelFormatType_32ARGB, options: nil).imageBufferValue!\n }\n\n func setIn_0(with in_0: URL) throws {\n self.in_0 = try MLFeatureValue(imageAt: in_0, pixelsWide: 320, pixelsHigh: 320, pixelFormatType: kCVPixelFormatType_32ARGB, options: nil).imageBufferValue!\n }\n\n}\n\n\n/// Model Prediction Output Type\n@available(macOS 11.0, iOS 14.0, tvOS 14.0, watchOS 7.0, *)\nclass segmentationOutput : MLFeatureProvider {\n\n /// Source provided by CoreML\n private let provider : MLFeatureProvider\n\n /// out_p0 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p0: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p0\")!.imageBufferValue\n }()!\n\n /// out_p1 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p1: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p1\")!.imageBufferValue\n }()!\n\n /// out_p2 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p2: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p2\")!.imageBufferValue\n }()!\n\n /// out_p3 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p3: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p3\")!.imageBufferValue\n }()!\n\n /// out_p4 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p4: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p4\")!.imageBufferValue\n }()!\n\n /// out_p5 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p5: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p5\")!.imageBufferValue\n }()!\n\n /// out_p6 as grayscale (kCVPixelFormatType_OneComponent8) image buffer, 320 pixels wide by 320 pixels high\n lazy var out_p6: CVPixelBuffer = {\n [unowned self] in return self.provider.featureValue(for: \"out_p6\")!.imageBufferValue\n }()!\n\n var featureNames: Set {\n return self.provider.featureNames\n }\n \n func featureValue(for featureName: String) -> MLFeatureValue? {\n return self.provider.featureValue(for: featureName)\n }\n\n init(out_p0: CVPixelBuffer, out_p1: CVPixelBuffer, out_p2: CVPixelBuffer, out_p3: CVPixelBuffer, out_p4: CVPixelBuffer, out_p5: CVPixelBuffer, out_p6: CVPixelBuffer) {\n self.provider = try! MLDictionaryFeatureProvider(dictionary: [\"out_p0\" : MLFeatureValue(pixelBuffer: out_p0), \"out_p1\" : MLFeatureValue(pixelBuffer: out_p1), \"out_p2\" : MLFeatureValue(pixelBuffer: out_p2), \"out_p3\" : MLFeatureValue(pixelBuffer: out_p3), \"out_p4\" : MLFeatureValue(pixelBuffer: out_p4), \"out_p5\" : MLFeatureValue(pixelBuffer: out_p5), \"out_p6\" : MLFeatureValue(pixelBuffer: out_p6)])\n }\n\n init(features: MLFeatureProvider) {\n self.provider = features\n }\n}\n\n\n/// Class for model loading and prediction\n@available(macOS 11.0, iOS 14.0, tvOS 14.0, watchOS 7.0, *)\nclass segmentation {\n let model: MLModel\n\n /// URL of model assuming it was installed in the same bundle as this class\n class var urlOfModelInThisBundle : URL {\n let bundle = Bundle(for: self)\n return bundle.url(forResource: \"segmentation\", withExtension:\"mlmodelc\")!\n }\n\n /**\n Construct segmentation instance with an existing MLModel object.\n\n Usually the application does not use this initializer unless it makes a subclass of segmentation.\n Such application may want to use `MLModel(contentsOfURL:configuration:)` and `segmentation.urlOfModelInThisBundle` to create a MLModel object to pass-in.\n\n - parameters:\n - model: MLModel object\n */\n init(model: MLModel) {\n self.model = model\n }\n\n /**\n Construct a model with configuration\n\n - parameters:\n - configuration: the desired model configuration\n\n - throws: an NSError object that describes the problem\n */\n convenience init(configuration: MLModelConfiguration = MLModelConfiguration()) throws {\n try self.init(contentsOf: type(of:self).urlOfModelInThisBundle, configuration: configuration)\n }\n\n /**\n Construct segmentation instance with explicit path to mlmodelc file\n - parameters:\n - modelURL: the file url of the model\n\n - throws: an NSError object that describes the problem\n */\n convenience init(contentsOf modelURL: URL) throws {\n try self.init(model: MLModel(contentsOf: modelURL))\n }\n\n /**\n Construct a model with URL of the .mlmodelc directory and configuration\n\n - parameters:\n - modelURL: the file url of the model\n - configuration: the desired model configuration\n\n - throws: an NSError object that describes the problem\n */\n convenience init(contentsOf modelURL: URL, configuration: MLModelConfiguration) throws {\n try self.init(model: MLModel(contentsOf: modelURL, configuration: configuration))\n }\n\n /**\n Construct segmentation instance asynchronously with optional configuration.\n\n Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.\n\n - parameters:\n - configuration: the desired model configuration\n - handler: the completion handler to be called when the model loading completes successfully or unsuccessfully\n */\n class func load(configuration: MLModelConfiguration = MLModelConfiguration(), completionHandler handler: @escaping (Swift.Result) -> Void) {\n return self.load(contentsOf: self.urlOfModelInThisBundle, configuration: configuration, completionHandler: handler)\n }\n\n /**\n Construct segmentation instance asynchronously with optional configuration.\n\n Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.\n\n - parameters:\n - configuration: the desired model configuration\n */\n @available(macOS 12.0, iOS 15.0, tvOS 15.0, watchOS 8.0, *)\n class func load(configuration: MLModelConfiguration = MLModelConfiguration()) async throws -> segmentation {\n return try await self.load(contentsOf: self.urlOfModelInThisBundle, configuration: configuration)\n }\n\n /**\n Construct segmentation instance asynchronously with URL of the .mlmodelc directory with optional configuration.\n\n Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.\n\n - parameters:\n - modelURL: the URL to the model\n - configuration: the desired model configuration\n - handler: the completion handler to be called when the model loading completes successfully or unsuccessfully\n */\n class func load(contentsOf modelURL: URL, configuration: MLModelConfiguration = MLModelConfiguration(), completionHandler handler: @escaping (Swift.Result) -> Void) {\n MLModel.load(contentsOf: modelURL, configuration: configuration) { result in\n switch result {\n case .failure(let error):\n handler(.failure(error))\n case .success(let model):\n handler(.success(segmentation(model: model)))\n }\n }\n }\n\n /**\n Construct segmentation instance asynchronously with URL of the .mlmodelc directory with optional configuration.\n\n Model loading may take time when the model content is not immediately available (e.g. encrypted model). Use this factory method especially when the caller is on the main thread.\n\n - parameters:\n - modelURL: the URL to the model\n - configuration: the desired model configuration\n */\n @available(macOS 12.0, iOS 15.0, tvOS 15.0, watchOS 8.0, *)\n class func load(contentsOf modelURL: URL, configuration: MLModelConfiguration = MLModelConfiguration()) async throws -> segmentation {\n let model = try await MLModel.load(contentsOf: modelURL, configuration: configuration)\n return segmentation(model: model)\n }\n\n /**\n Make a prediction using the structured interface\n\n - parameters:\n - input: the input to the prediction as segmentationInput\n\n - throws: an NSError object that describes the problem\n\n - returns: the result of the prediction as segmentationOutput\n */\n func prediction(input: segmentationInput) throws -> segmentationOutput {\n return try self.prediction(input: input, options: MLPredictionOptions())\n }\n\n /**\n Make a prediction using the structured interface\n\n - parameters:\n - input: the input to the prediction as segmentationInput\n - options: prediction options \n\n - throws: an NSError object that describes the problem\n\n - returns: the result of the prediction as segmentationOutput\n */\n func prediction(input: segmentationInput, options: MLPredictionOptions) throws -> segmentationOutput {\n let outFeatures = try model.prediction(from: input, options:options)\n return segmentationOutput(features: outFeatures)\n }\n\n /**\n Make a prediction using the convenience interface\n\n - parameters:\n - in_0 as color (kCVPixelFormatType_32BGRA) image buffer, 320 pixels wide by 320 pixels high\n\n - throws: an NSError object that describes the problem\n\n - returns: the result of the prediction as segmentationOutput\n */\n func prediction(in_0: CVPixelBuffer) throws -> segmentationOutput {\n let input_ = segmentationInput(in_0: in_0)\n return try self.prediction(input: input_)\n }\n\n /**\n Make a batch prediction using the structured interface\n\n - parameters:\n - inputs: the inputs to the prediction as [segmentationInput]\n - options: prediction options \n\n - throws: an NSError object that describes the problem\n\n - returns: the result of the prediction as [segmentationOutput]\n */\n func predictions(inputs: [segmentationInput], options: MLPredictionOptions = MLPredictionOptions()) throws -> [segmentationOutput] {\n let batchIn = MLArrayBatchProvider(array: inputs)\n let batchOut = try model.predictions(from: batchIn, options: options)\n var results : [segmentationOutput] = []\n results.reserveCapacity(inputs.count)\n for i in 0..