OpenCV Computer Vision AI Skill Guide

Overview & Engine Architecture

OpenCV (cv2) provides image/video IO, geometric and photometric transforms, classical feature detectors, and drawing utilities. Images are NumPy arrays in BGR order by default. Agents track color-space conversions explicitly, keep resize/crop parameters reproducible, release video captures, and push deep-learning detection to @pytorch / @huggingface-transformers when classical methods plateau.

imread / VideoCapture
      -> BGR ndarray
          -> cvtColor / resize / filter / threshold
          -> contours / features / write

When to use this skill

• Preprocessing images for ML models
• Classical detection (edges, contours, template match)
• Frame sampling and annotation from video