{
  "schemaVersion": "1.0",
  "item": {
    "slug": "senior-computer-vision",
    "name": "Senior Computer Vision",
    "source": "tencent",
    "type": "skill",
    "category": "AI 智能",
    "sourceUrl": "https://clawhub.ai/alirezarezvani/senior-computer-vision",
    "canonicalUrl": "https://clawhub.ai/alirezarezvani/senior-computer-vision",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/senior-computer-vision",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=senior-computer-vision",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "SKILL.md",
      "references/computer_vision_architectures.md",
      "references/object_detection_optimization.md",
      "references/production_vision_systems.md",
      "references/reference-docs-and-commands.md",
      "scripts/dataset_pipeline_builder.py"
    ],
    "primaryDoc": "SKILL.md",
    "quickSetup": [
      "Download the package from Yavira.",
      "Extract the archive and review SKILL.md first.",
      "Import or place the package into your OpenClaw setup."
    ],
    "agentAssist": {
      "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
      "steps": [
        "Download the package from Yavira.",
        "Extract it into a folder your agent can access.",
        "Paste one of the prompts below and point your agent at the extracted folder."
      ],
      "prompts": [
        {
          "label": "New install",
          "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
        },
        {
          "label": "Upgrade existing",
          "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
        }
      ]
    },
    "sourceHealth": {
      "source": "tencent",
      "status": "healthy",
      "reason": "direct_download_ok",
      "recommendedAction": "download",
      "checkedAt": "2026-04-30T16:55:25.780Z",
      "expiresAt": "2026-05-07T16:55:25.780Z",
      "httpStatus": 200,
      "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
      "contentType": "application/zip",
      "probeMethod": "head",
      "details": {
        "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
        "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"",
        "redirectLocation": null,
        "bodySnippet": null
      },
      "scope": "source",
      "summary": "Source download looks usable.",
      "detail": "Yavira can redirect you to the upstream package for this source.",
      "primaryActionLabel": "Download for OpenClaw",
      "primaryActionHref": "/downloads/senior-computer-vision"
    },
    "validation": {
      "installChecklist": [
        "Use the Yavira download entry.",
        "Review SKILL.md after the package is downloaded.",
        "Confirm the extracted package contains the expected setup assets."
      ],
      "postInstallChecks": [
        "Confirm the extracted package includes the expected docs or setup files.",
        "Validate the skill or prompts are available in your target agent workspace.",
        "Capture any manual follow-up steps the agent could not complete."
      ]
    },
    "downloadPageUrl": "https://openagent3.xyz/downloads/senior-computer-vision",
    "agentPageUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent",
    "manifestUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent.md"
  },
  "agentAssist": {
    "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
    "steps": [
      "Download the package from Yavira.",
      "Extract it into a folder your agent can access.",
      "Paste one of the prompts below and point your agent at the extracted folder."
    ],
    "prompts": [
      {
        "label": "New install",
        "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
      },
      {
        "label": "Upgrade existing",
        "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
      }
    ]
  },
  "documentation": {
    "source": "clawhub",
    "primaryDoc": "SKILL.md",
    "sections": [
      {
        "title": "Senior Computer Vision Engineer",
        "body": "Production computer vision engineering skill for object detection, image segmentation, and visual AI system deployment."
      },
      {
        "title": "Table of Contents",
        "body": "Quick Start\nCore Expertise\nTech Stack\nWorkflow 1: Object Detection Pipeline\nWorkflow 2: Model Optimization and Deployment\nWorkflow 3: Custom Dataset Preparation\nArchitecture Selection Guide\nReference Documentation\nCommon Commands"
      },
      {
        "title": "Quick Start",
        "body": "# Generate training configuration for YOLO or Faster R-CNN\npython scripts/vision_model_trainer.py models/ --task detection --arch yolov8\n\n# Analyze model for optimization opportunities (quantization, pruning)\npython scripts/inference_optimizer.py model.pt --target onnx --benchmark\n\n# Build dataset pipeline with augmentations\npython scripts/dataset_pipeline_builder.py images/ --format coco --augment"
      },
      {
        "title": "Core Expertise",
        "body": "This skill provides guidance on:\n\nObject Detection: YOLO family (v5-v11), Faster R-CNN, DETR, RT-DETR\nInstance Segmentation: Mask R-CNN, YOLACT, SOLOv2\nSemantic Segmentation: DeepLabV3+, SegFormer, SAM (Segment Anything)\nImage Classification: ResNet, EfficientNet, Vision Transformers (ViT, DeiT)\nVideo Analysis: Object tracking (ByteTrack, SORT), action recognition\n3D Vision: Depth estimation, point cloud processing, NeRF\nProduction Deployment: ONNX, TensorRT, OpenVINO, CoreML"
      },
      {
        "title": "Tech Stack",
        "body": "CategoryTechnologiesFrameworksPyTorch, torchvision, timmDetectionUltralytics (YOLO), Detectron2, MMDetectionSegmentationsegment-anything, mmsegmentationOptimizationONNX, TensorRT, OpenVINO, torch.compileImage ProcessingOpenCV, Pillow, albumentationsAnnotationCVAT, Label Studio, RoboflowExperiment TrackingMLflow, Weights & BiasesServingTriton Inference Server, TorchServe"
      },
      {
        "title": "Workflow 1: Object Detection Pipeline",
        "body": "Use this workflow when building an object detection system from scratch."
      },
      {
        "title": "Step 1: Define Detection Requirements",
        "body": "Analyze the detection task requirements:\n\nDetection Requirements Analysis:\n- Target objects: [list specific classes to detect]\n- Real-time requirement: [yes/no, target FPS]\n- Accuracy priority: [speed vs accuracy trade-off]\n- Deployment target: [cloud GPU, edge device, mobile]\n- Dataset size: [number of images, annotations per class]"
      },
      {
        "title": "Step 2: Select Detection Architecture",
        "body": "Choose architecture based on requirements:\n\nRequirementRecommended ArchitectureWhyReal-time (>30 FPS)YOLOv8/v11, RT-DETRSingle-stage, optimized for speedHigh accuracyFaster R-CNN, DINOTwo-stage, better localizationSmall objectsYOLO + SAHI, Faster R-CNN + FPNMulti-scale detectionEdge deploymentYOLOv8n, MobileNetV3-SSDLightweight architecturesTransformer-basedDETR, DINO, RT-DETREnd-to-end, no NMS required"
      },
      {
        "title": "Step 3: Prepare Dataset",
        "body": "Convert annotations to required format:\n\n# COCO format (recommended)\npython scripts/dataset_pipeline_builder.py data/images/ \\\n    --annotations data/labels/ \\\n    --format coco \\\n    --split 0.8 0.1 0.1 \\\n    --output data/coco/\n\n# Verify dataset\npython -c \"from pycocotools.coco import COCO; coco = COCO('data/coco/train.json'); print(f'Images: {len(coco.imgs)}, Categories: {len(coco.cats)}')\""
      },
      {
        "title": "Step 4: Configure Training",
        "body": "Generate training configuration:\n\n# For Ultralytics YOLO\npython scripts/vision_model_trainer.py data/coco/ \\\n    --task detection \\\n    --arch yolov8m \\\n    --epochs 100 \\\n    --batch 16 \\\n    --imgsz 640 \\\n    --output configs/\n\n# For Detectron2\npython scripts/vision_model_trainer.py data/coco/ \\\n    --task detection \\\n    --arch faster_rcnn_R_50_FPN \\\n    --framework detectron2 \\\n    --output configs/"
      },
      {
        "title": "Step 5: Train and Validate",
        "body": "# Ultralytics training\nyolo detect train data=data.yaml model=yolov8m.pt epochs=100 imgsz=640\n\n# Detectron2 training\npython train_net.py --config-file configs/faster_rcnn.yaml --num-gpus 1\n\n# Validate on test set\nyolo detect val model=runs/detect/train/weights/best.pt data=data.yaml"
      },
      {
        "title": "Step 6: Evaluate Results",
        "body": "Key metrics to analyze:\n\nMetricTargetDescriptionmAP@50>0.7Mean Average Precision at IoU 0.5mAP@50:95>0.5COCO primary metricPrecision>0.8Low false positivesRecall>0.8Low missed detectionsInference time<33msFor 30 FPS real-time"
      },
      {
        "title": "Workflow 2: Model Optimization and Deployment",
        "body": "Use this workflow when preparing a trained model for production deployment."
      },
      {
        "title": "Step 1: Benchmark Baseline Performance",
        "body": "# Measure current model performance\npython scripts/inference_optimizer.py model.pt \\\n    --benchmark \\\n    --input-size 640 640 \\\n    --batch-sizes 1 4 8 16 \\\n    --warmup 10 \\\n    --iterations 100\n\nExpected output:\n\nBaseline Performance (PyTorch FP32):\n- Batch 1: 45.2ms (22.1 FPS)\n- Batch 4: 89.4ms (44.7 FPS)\n- Batch 8: 165.3ms (48.4 FPS)\n- Memory: 2.1 GB\n- Parameters: 25.9M"
      },
      {
        "title": "Step 2: Select Optimization Strategy",
        "body": "Deployment TargetOptimization PathNVIDIA GPU (cloud)PyTorch → ONNX → TensorRT FP16NVIDIA GPU (edge)PyTorch → TensorRT INT8Intel CPUPyTorch → ONNX → OpenVINOApple SiliconPyTorch → CoreMLGeneric CPUPyTorch → ONNX RuntimeMobilePyTorch → TFLite or ONNX Mobile"
      },
      {
        "title": "Step 3: Export to ONNX",
        "body": "# Export with dynamic batch size\npython scripts/inference_optimizer.py model.pt \\\n    --export onnx \\\n    --input-size 640 640 \\\n    --dynamic-batch \\\n    --simplify \\\n    --output model.onnx\n\n# Verify ONNX model\npython -c \"import onnx; model = onnx.load('model.onnx'); onnx.checker.check_model(model); print('ONNX model valid')\""
      },
      {
        "title": "Step 4: Apply Quantization (Optional)",
        "body": "For INT8 quantization with calibration:\n\n# Generate calibration dataset\npython scripts/inference_optimizer.py model.onnx \\\n    --quantize int8 \\\n    --calibration-data data/calibration/ \\\n    --calibration-samples 500 \\\n    --output model_int8.onnx\n\nQuantization impact analysis:\n\nPrecisionSizeSpeedAccuracy DropFP32100%1x0%FP1650%1.5-2x<0.5%INT825%2-4x1-3%"
      },
      {
        "title": "Step 5: Convert to Target Runtime",
        "body": "# TensorRT (NVIDIA GPU)\ntrtexec --onnx=model.onnx --saveEngine=model.engine --fp16\n\n# OpenVINO (Intel)\nmo --input_model model.onnx --output_dir openvino/\n\n# CoreML (Apple)\npython -c \"import coremltools as ct; model = ct.convert('model.onnx'); model.save('model.mlpackage')\""
      },
      {
        "title": "Step 6: Benchmark Optimized Model",
        "body": "python scripts/inference_optimizer.py model.engine \\\n    --benchmark \\\n    --runtime tensorrt \\\n    --compare model.pt\n\nExpected speedup:\n\nOptimization Results:\n- Original (PyTorch FP32): 45.2ms\n- Optimized (TensorRT FP16): 12.8ms\n- Speedup: 3.5x\n- Accuracy change: -0.3% mAP"
      },
      {
        "title": "Workflow 3: Custom Dataset Preparation",
        "body": "Use this workflow when preparing a computer vision dataset for training."
      },
      {
        "title": "Step 1: Audit Raw Data",
        "body": "# Analyze image dataset\npython scripts/dataset_pipeline_builder.py data/raw/ \\\n    --analyze \\\n    --output analysis/\n\nAnalysis report includes:\n\nDataset Analysis:\n- Total images: 5,234\n- Image sizes: 640x480 to 4096x3072 (variable)\n- Formats: JPEG (4,891), PNG (343)\n- Corrupted: 12 files\n- Duplicates: 45 pairs\n\nAnnotation Analysis:\n- Format detected: Pascal VOC XML\n- Total annotations: 28,456\n- Classes: 5 (car, person, bicycle, dog, cat)\n- Distribution: car (12,340), person (8,234), bicycle (3,456), dog (2,890), cat (1,536)\n- Empty images: 234"
      },
      {
        "title": "Step 2: Clean and Validate",
        "body": "# Remove corrupted and duplicate images\npython scripts/dataset_pipeline_builder.py data/raw/ \\\n    --clean \\\n    --remove-corrupted \\\n    --remove-duplicates \\\n    --output data/cleaned/"
      },
      {
        "title": "Step 3: Convert Annotation Format",
        "body": "# Convert VOC to COCO format\npython scripts/dataset_pipeline_builder.py data/cleaned/ \\\n    --annotations data/annotations/ \\\n    --input-format voc \\\n    --output-format coco \\\n    --output data/coco/\n\nSupported format conversions:\n\nFromToPascal VOC XMLCOCO JSONYOLO TXTCOCO JSONCOCO JSONYOLO TXTLabelMe JSONCOCO JSONCVAT XMLCOCO JSON"
      },
      {
        "title": "Step 4: Apply Augmentations",
        "body": "# Generate augmentation config\npython scripts/dataset_pipeline_builder.py data/coco/ \\\n    --augment \\\n    --aug-config configs/augmentation.yaml \\\n    --output data/augmented/\n\nRecommended augmentations for detection:\n\n# configs/augmentation.yaml\naugmentations:\n  geometric:\n    - horizontal_flip: { p: 0.5 }\n    - vertical_flip: { p: 0.1 }  # Only if orientation invariant\n    - rotate: { limit: 15, p: 0.3 }\n    - scale: { scale_limit: 0.2, p: 0.5 }\n\n  color:\n    - brightness_contrast: { brightness_limit: 0.2, contrast_limit: 0.2, p: 0.5 }\n    - hue_saturation: { hue_shift_limit: 20, sat_shift_limit: 30, p: 0.3 }\n    - blur: { blur_limit: 3, p: 0.1 }\n\n  advanced:\n    - mosaic: { p: 0.5 }  # YOLO-style mosaic\n    - mixup: { p: 0.1 }   # Image mixing\n    - cutout: { num_holes: 8, max_h_size: 32, max_w_size: 32, p: 0.3 }"
      },
      {
        "title": "Step 5: Create Train/Val/Test Splits",
        "body": "python scripts/dataset_pipeline_builder.py data/augmented/ \\\n    --split 0.8 0.1 0.1 \\\n    --stratify \\\n    --seed 42 \\\n    --output data/final/\n\nSplit strategy guidelines:\n\nDataset SizeTrainValTest<1,000 images70%15%15%1,000-10,00080%10%10%>10,00090%5%5%"
      },
      {
        "title": "Step 6: Generate Dataset Configuration",
        "body": "# For Ultralytics YOLO\npython scripts/dataset_pipeline_builder.py data/final/ \\\n    --generate-config yolo \\\n    --output data.yaml\n\n# For Detectron2\npython scripts/dataset_pipeline_builder.py data/final/ \\\n    --generate-config detectron2 \\\n    --output detectron2_config.py"
      },
      {
        "title": "Object Detection Architectures",
        "body": "ArchitectureSpeedAccuracyBest ForYOLOv8n1.2ms37.3 mAPEdge, mobile, real-timeYOLOv8s2.1ms44.9 mAPBalanced speed/accuracyYOLOv8m4.2ms50.2 mAPGeneral purposeYOLOv8l6.8ms52.9 mAPHigh accuracyYOLOv8x10.1ms53.9 mAPMaximum accuracyRT-DETR-L5.3ms53.0 mAPTransformer, no NMSFaster R-CNN R5046ms40.2 mAPTwo-stage, high qualityDINO-4scale85ms49.0 mAPSOTA transformer"
      },
      {
        "title": "Segmentation Architectures",
        "body": "ArchitectureTypeSpeedBest ForYOLOv8-segInstance4.5msReal-time instance segMask R-CNNInstance67msHigh-quality masksSAMPromptable50msZero-shot segmentationDeepLabV3+Semantic25msScene parsingSegFormerSemantic15msEfficient semantic seg"
      },
      {
        "title": "CNN vs Vision Transformer Trade-offs",
        "body": "AspectCNN (YOLO, R-CNN)ViT (DETR, DINO)Training data needed1K-10K images10K-100K+ imagesTraining timeFastSlow (needs more epochs)Inference speedFasterSlowerSmall objectsGood with FPNNeeds multi-scaleGlobal contextLimitedExcellentPositional encodingImplicitExplicit"
      },
      {
        "title": "Reference Documentation",
        "body": "→ See references/reference-docs-and-commands.md for details"
      },
      {
        "title": "Performance Targets",
        "body": "MetricReal-timeHigh AccuracyEdgeFPS>30>10>15mAP@50>0.6>0.8>0.5Latency P99<50ms<150ms<100msGPU Memory<4GB<8GB<2GBModel Size<50MB<200MB<20MB"
      },
      {
        "title": "Resources",
        "body": "Architecture Guide: references/computer_vision_architectures.md\nOptimization Guide: references/object_detection_optimization.md\nDeployment Guide: references/production_vision_systems.md\nScripts: scripts/ directory for automation tools"
      }
    ],
    "body": "Senior Computer Vision Engineer\n\nProduction computer vision engineering skill for object detection, image segmentation, and visual AI system deployment.\n\nTable of Contents\nQuick Start\nCore Expertise\nTech Stack\nWorkflow 1: Object Detection Pipeline\nWorkflow 2: Model Optimization and Deployment\nWorkflow 3: Custom Dataset Preparation\nArchitecture Selection Guide\nReference Documentation\nCommon Commands\nQuick Start\n# Generate training configuration for YOLO or Faster R-CNN\npython scripts/vision_model_trainer.py models/ --task detection --arch yolov8\n\n# Analyze model for optimization opportunities (quantization, pruning)\npython scripts/inference_optimizer.py model.pt --target onnx --benchmark\n\n# Build dataset pipeline with augmentations\npython scripts/dataset_pipeline_builder.py images/ --format coco --augment\n\nCore Expertise\n\nThis skill provides guidance on:\n\nObject Detection: YOLO family (v5-v11), Faster R-CNN, DETR, RT-DETR\nInstance Segmentation: Mask R-CNN, YOLACT, SOLOv2\nSemantic Segmentation: DeepLabV3+, SegFormer, SAM (Segment Anything)\nImage Classification: ResNet, EfficientNet, Vision Transformers (ViT, DeiT)\nVideo Analysis: Object tracking (ByteTrack, SORT), action recognition\n3D Vision: Depth estimation, point cloud processing, NeRF\nProduction Deployment: ONNX, TensorRT, OpenVINO, CoreML\nTech Stack\nCategory\tTechnologies\nFrameworks\tPyTorch, torchvision, timm\nDetection\tUltralytics (YOLO), Detectron2, MMDetection\nSegmentation\tsegment-anything, mmsegmentation\nOptimization\tONNX, TensorRT, OpenVINO, torch.compile\nImage Processing\tOpenCV, Pillow, albumentations\nAnnotation\tCVAT, Label Studio, Roboflow\nExperiment Tracking\tMLflow, Weights & Biases\nServing\tTriton Inference Server, TorchServe\nWorkflow 1: Object Detection Pipeline\n\nUse this workflow when building an object detection system from scratch.\n\nStep 1: Define Detection Requirements\n\nAnalyze the detection task requirements:\n\nDetection Requirements Analysis:\n- Target objects: [list specific classes to detect]\n- Real-time requirement: [yes/no, target FPS]\n- Accuracy priority: [speed vs accuracy trade-off]\n- Deployment target: [cloud GPU, edge device, mobile]\n- Dataset size: [number of images, annotations per class]\n\nStep 2: Select Detection Architecture\n\nChoose architecture based on requirements:\n\nRequirement\tRecommended Architecture\tWhy\nReal-time (>30 FPS)\tYOLOv8/v11, RT-DETR\tSingle-stage, optimized for speed\nHigh accuracy\tFaster R-CNN, DINO\tTwo-stage, better localization\nSmall objects\tYOLO + SAHI, Faster R-CNN + FPN\tMulti-scale detection\nEdge deployment\tYOLOv8n, MobileNetV3-SSD\tLightweight architectures\nTransformer-based\tDETR, DINO, RT-DETR\tEnd-to-end, no NMS required\nStep 3: Prepare Dataset\n\nConvert annotations to required format:\n\n# COCO format (recommended)\npython scripts/dataset_pipeline_builder.py data/images/ \\\n    --annotations data/labels/ \\\n    --format coco \\\n    --split 0.8 0.1 0.1 \\\n    --output data/coco/\n\n# Verify dataset\npython -c \"from pycocotools.coco import COCO; coco = COCO('data/coco/train.json'); print(f'Images: {len(coco.imgs)}, Categories: {len(coco.cats)}')\"\n\nStep 4: Configure Training\n\nGenerate training configuration:\n\n# For Ultralytics YOLO\npython scripts/vision_model_trainer.py data/coco/ \\\n    --task detection \\\n    --arch yolov8m \\\n    --epochs 100 \\\n    --batch 16 \\\n    --imgsz 640 \\\n    --output configs/\n\n# For Detectron2\npython scripts/vision_model_trainer.py data/coco/ \\\n    --task detection \\\n    --arch faster_rcnn_R_50_FPN \\\n    --framework detectron2 \\\n    --output configs/\n\nStep 5: Train and Validate\n# Ultralytics training\nyolo detect train data=data.yaml model=yolov8m.pt epochs=100 imgsz=640\n\n# Detectron2 training\npython train_net.py --config-file configs/faster_rcnn.yaml --num-gpus 1\n\n# Validate on test set\nyolo detect val model=runs/detect/train/weights/best.pt data=data.yaml\n\nStep 6: Evaluate Results\n\nKey metrics to analyze:\n\nMetric\tTarget\tDescription\nmAP@50\t>0.7\tMean Average Precision at IoU 0.5\nmAP@50:95\t>0.5\tCOCO primary metric\nPrecision\t>0.8\tLow false positives\nRecall\t>0.8\tLow missed detections\nInference time\t<33ms\tFor 30 FPS real-time\nWorkflow 2: Model Optimization and Deployment\n\nUse this workflow when preparing a trained model for production deployment.\n\nStep 1: Benchmark Baseline Performance\n# Measure current model performance\npython scripts/inference_optimizer.py model.pt \\\n    --benchmark \\\n    --input-size 640 640 \\\n    --batch-sizes 1 4 8 16 \\\n    --warmup 10 \\\n    --iterations 100\n\n\nExpected output:\n\nBaseline Performance (PyTorch FP32):\n- Batch 1: 45.2ms (22.1 FPS)\n- Batch 4: 89.4ms (44.7 FPS)\n- Batch 8: 165.3ms (48.4 FPS)\n- Memory: 2.1 GB\n- Parameters: 25.9M\n\nStep 2: Select Optimization Strategy\nDeployment Target\tOptimization Path\nNVIDIA GPU (cloud)\tPyTorch → ONNX → TensorRT FP16\nNVIDIA GPU (edge)\tPyTorch → TensorRT INT8\nIntel CPU\tPyTorch → ONNX → OpenVINO\nApple Silicon\tPyTorch → CoreML\nGeneric CPU\tPyTorch → ONNX Runtime\nMobile\tPyTorch → TFLite or ONNX Mobile\nStep 3: Export to ONNX\n# Export with dynamic batch size\npython scripts/inference_optimizer.py model.pt \\\n    --export onnx \\\n    --input-size 640 640 \\\n    --dynamic-batch \\\n    --simplify \\\n    --output model.onnx\n\n# Verify ONNX model\npython -c \"import onnx; model = onnx.load('model.onnx'); onnx.checker.check_model(model); print('ONNX model valid')\"\n\nStep 4: Apply Quantization (Optional)\n\nFor INT8 quantization with calibration:\n\n# Generate calibration dataset\npython scripts/inference_optimizer.py model.onnx \\\n    --quantize int8 \\\n    --calibration-data data/calibration/ \\\n    --calibration-samples 500 \\\n    --output model_int8.onnx\n\n\nQuantization impact analysis:\n\nPrecision\tSize\tSpeed\tAccuracy Drop\nFP32\t100%\t1x\t0%\nFP16\t50%\t1.5-2x\t<0.5%\nINT8\t25%\t2-4x\t1-3%\nStep 5: Convert to Target Runtime\n# TensorRT (NVIDIA GPU)\ntrtexec --onnx=model.onnx --saveEngine=model.engine --fp16\n\n# OpenVINO (Intel)\nmo --input_model model.onnx --output_dir openvino/\n\n# CoreML (Apple)\npython -c \"import coremltools as ct; model = ct.convert('model.onnx'); model.save('model.mlpackage')\"\n\nStep 6: Benchmark Optimized Model\npython scripts/inference_optimizer.py model.engine \\\n    --benchmark \\\n    --runtime tensorrt \\\n    --compare model.pt\n\n\nExpected speedup:\n\nOptimization Results:\n- Original (PyTorch FP32): 45.2ms\n- Optimized (TensorRT FP16): 12.8ms\n- Speedup: 3.5x\n- Accuracy change: -0.3% mAP\n\nWorkflow 3: Custom Dataset Preparation\n\nUse this workflow when preparing a computer vision dataset for training.\n\nStep 1: Audit Raw Data\n# Analyze image dataset\npython scripts/dataset_pipeline_builder.py data/raw/ \\\n    --analyze \\\n    --output analysis/\n\n\nAnalysis report includes:\n\nDataset Analysis:\n- Total images: 5,234\n- Image sizes: 640x480 to 4096x3072 (variable)\n- Formats: JPEG (4,891), PNG (343)\n- Corrupted: 12 files\n- Duplicates: 45 pairs\n\nAnnotation Analysis:\n- Format detected: Pascal VOC XML\n- Total annotations: 28,456\n- Classes: 5 (car, person, bicycle, dog, cat)\n- Distribution: car (12,340), person (8,234), bicycle (3,456), dog (2,890), cat (1,536)\n- Empty images: 234\n\nStep 2: Clean and Validate\n# Remove corrupted and duplicate images\npython scripts/dataset_pipeline_builder.py data/raw/ \\\n    --clean \\\n    --remove-corrupted \\\n    --remove-duplicates \\\n    --output data/cleaned/\n\nStep 3: Convert Annotation Format\n# Convert VOC to COCO format\npython scripts/dataset_pipeline_builder.py data/cleaned/ \\\n    --annotations data/annotations/ \\\n    --input-format voc \\\n    --output-format coco \\\n    --output data/coco/\n\n\nSupported format conversions:\n\nFrom\tTo\nPascal VOC XML\tCOCO JSON\nYOLO TXT\tCOCO JSON\nCOCO JSON\tYOLO TXT\nLabelMe JSON\tCOCO JSON\nCVAT XML\tCOCO JSON\nStep 4: Apply Augmentations\n# Generate augmentation config\npython scripts/dataset_pipeline_builder.py data/coco/ \\\n    --augment \\\n    --aug-config configs/augmentation.yaml \\\n    --output data/augmented/\n\n\nRecommended augmentations for detection:\n\n# configs/augmentation.yaml\naugmentations:\n  geometric:\n    - horizontal_flip: { p: 0.5 }\n    - vertical_flip: { p: 0.1 }  # Only if orientation invariant\n    - rotate: { limit: 15, p: 0.3 }\n    - scale: { scale_limit: 0.2, p: 0.5 }\n\n  color:\n    - brightness_contrast: { brightness_limit: 0.2, contrast_limit: 0.2, p: 0.5 }\n    - hue_saturation: { hue_shift_limit: 20, sat_shift_limit: 30, p: 0.3 }\n    - blur: { blur_limit: 3, p: 0.1 }\n\n  advanced:\n    - mosaic: { p: 0.5 }  # YOLO-style mosaic\n    - mixup: { p: 0.1 }   # Image mixing\n    - cutout: { num_holes: 8, max_h_size: 32, max_w_size: 32, p: 0.3 }\n\nStep 5: Create Train/Val/Test Splits\npython scripts/dataset_pipeline_builder.py data/augmented/ \\\n    --split 0.8 0.1 0.1 \\\n    --stratify \\\n    --seed 42 \\\n    --output data/final/\n\n\nSplit strategy guidelines:\n\nDataset Size\tTrain\tVal\tTest\n<1,000 images\t70%\t15%\t15%\n1,000-10,000\t80%\t10%\t10%\n>10,000\t90%\t5%\t5%\nStep 6: Generate Dataset Configuration\n# For Ultralytics YOLO\npython scripts/dataset_pipeline_builder.py data/final/ \\\n    --generate-config yolo \\\n    --output data.yaml\n\n# For Detectron2\npython scripts/dataset_pipeline_builder.py data/final/ \\\n    --generate-config detectron2 \\\n    --output detectron2_config.py\n\nArchitecture Selection Guide\nObject Detection Architectures\nArchitecture\tSpeed\tAccuracy\tBest For\nYOLOv8n\t1.2ms\t37.3 mAP\tEdge, mobile, real-time\nYOLOv8s\t2.1ms\t44.9 mAP\tBalanced speed/accuracy\nYOLOv8m\t4.2ms\t50.2 mAP\tGeneral purpose\nYOLOv8l\t6.8ms\t52.9 mAP\tHigh accuracy\nYOLOv8x\t10.1ms\t53.9 mAP\tMaximum accuracy\nRT-DETR-L\t5.3ms\t53.0 mAP\tTransformer, no NMS\nFaster R-CNN R50\t46ms\t40.2 mAP\tTwo-stage, high quality\nDINO-4scale\t85ms\t49.0 mAP\tSOTA transformer\nSegmentation Architectures\nArchitecture\tType\tSpeed\tBest For\nYOLOv8-seg\tInstance\t4.5ms\tReal-time instance seg\nMask R-CNN\tInstance\t67ms\tHigh-quality masks\nSAM\tPromptable\t50ms\tZero-shot segmentation\nDeepLabV3+\tSemantic\t25ms\tScene parsing\nSegFormer\tSemantic\t15ms\tEfficient semantic seg\nCNN vs Vision Transformer Trade-offs\nAspect\tCNN (YOLO, R-CNN)\tViT (DETR, DINO)\nTraining data needed\t1K-10K images\t10K-100K+ images\nTraining time\tFast\tSlow (needs more epochs)\nInference speed\tFaster\tSlower\nSmall objects\tGood with FPN\tNeeds multi-scale\nGlobal context\tLimited\tExcellent\nPositional encoding\tImplicit\tExplicit\nReference Documentation\n\n→ See references/reference-docs-and-commands.md for details\n\nPerformance Targets\nMetric\tReal-time\tHigh Accuracy\tEdge\nFPS\t>30\t>10\t>15\nmAP@50\t>0.6\t>0.8\t>0.5\nLatency P99\t<50ms\t<150ms\t<100ms\nGPU Memory\t<4GB\t<8GB\t<2GB\nModel Size\t<50MB\t<200MB\t<20MB\nResources\nArchitecture Guide: references/computer_vision_architectures.md\nOptimization Guide: references/object_detection_optimization.md\nDeployment Guide: references/production_vision_systems.md\nScripts: scripts/ directory for automation tools"
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/alirezarezvani/senior-computer-vision",
    "publisherUrl": "https://clawhub.ai/alirezarezvani/senior-computer-vision",
    "owner": "alirezarezvani",
    "version": "2.1.1",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/senior-computer-vision",
    "downloadUrl": "https://openagent3.xyz/downloads/senior-computer-vision",
    "agentUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent",
    "manifestUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/senior-computer-vision/agent.md"
  }
}