Builds production computer vision pipelines for object detection and tracking in video, with real guidance on preprocessing frames, frame sampling strategies, and model selection between YOLO, Detectron2, and others. The anti-patterns section is especially useful, showing concrete examples of common mistakes like processing every frame in a 30 FPS video instead of sampling intelligently or detecting scene changes. Designed for drone footage analysis, wildlife monitoring, and real-time detection systems where you need actual pipeline code, not just API calls. Pairs well with geospatial skills if you're working with drone coordinates or data pipeline skills for handling video ETL.
npx skills add https://github.com/erichowens/some_claude_skills --skill computer-vision-pipeline