Computer Vision skills for AI agents

Expert guidance for building computer vision datasets, annotation workflows, data augmentation, and quality assurance. Covers annotation tools, synthetic data, dataset formats, versioning, and active learning loops. Trigger: dataset, annotation, labeling, data collection, CVAT, Label Studio, Roboflow, data augmentation, albumentations, synthetic data, dataset format, COCO format, YOLO format.

Expert guidance for deploying computer vision models on edge devices. Covers model optimization, quantization, TensorRT, OpenVINO, TFLite, ONNX, Jetson, and complete edge CV pipelines. Trigger: edge deployment, model optimization, quantization, TensorRT, ONNX, OpenVINO, Jetson, edge inference, model compression, TFLite, deploy model on device, real-time inference.

Expert guidance for face detection, recognition, alignment, and analysis systems. Covers ArcFace, InsightFace, DeepFace, face embeddings, anti-spoofing, and ethical deployment. Trigger: face recognition, face detection, facial recognition, face ID, face verification, face embedding, ArcFace, InsightFace, face matching.

Expert guidance for generative image and video models including diffusion models, Stable Diffusion, ControlNet, GANs, super-resolution, and fine-tuning techniques. Trigger: image generation, Stable Diffusion, diffusion model, ControlNet, DreamBooth, LoRA, text-to-image, inpainting, super-resolution, GAN, generative AI images.

Expert guidance for building image classification pipelines with deep learning. Covers CNN architectures, Vision Transformers, transfer learning, training loops, evaluation metrics, and deployment. Trigger: image classification, classify images, CNN, ResNet, EfficientNet, ViT, transfer learning, fine-tune image model.

Expert guidance for semantic, instance, and panoptic segmentation. Covers U-Net, DeepLab, Mask R-CNN, SAM/SAM2, loss functions, evaluation metrics, and domain-specific strategies. Trigger: image segmentation, semantic segmentation, instance segmentation, pixel-level, mask, U-Net, SAM, Segment Anything, panoptic.

Expert guidance for building object detection systems. Covers YOLO family, Faster R-CNN, DETR, Ultralytics ecosystem, custom training, evaluation metrics, and real-time deployment. Trigger: object detection, YOLO, bounding box, detect objects, Faster R-CNN, DETR, mAP, Ultralytics.

Expert guidance for OCR, text detection/recognition, and document understanding pipelines. Covers Tesseract, EasyOCR, PaddleOCR, document layout analysis, table extraction, and LLM-powered document parsing. Trigger: OCR, text recognition, document parsing, text extraction, PDF extraction, receipt parsing, invoice, handwriting recognition, document understanding.

Expert guidance for human pose estimation, body tracking, and gesture recognition. Covers OpenPose, MediaPipe, HRNet, ViTPose, 3D pose, hand tracking, and applications in fitness, sports, and AR. Trigger: pose estimation, body tracking, skeleton detection, keypoint detection, hand tracking, gesture recognition, MediaPipe Pose, OpenPose, human pose, body keypoints.

Expert guidance for video understanding, object tracking, action recognition, and real-time video processing. Covers SORT, DeepSORT, ByteTrack, SlowFast, optical flow, and scalable video pipelines. Trigger: video analysis, object tracking, multi-object tracking, action recognition, video classification, optical flow, surveillance, video processing, MOT.