Video Analytics System — Real-Time Object Tracking & Activity

I will build a video analytics system — combining object detection, multi-object tracking across frames, activity recognition, and configurable alert triggers for specific events (person entering a zone, count exceeding a threshold, or a defined activity occurring) — processing video streams in real-time or from recorded footage. Video analytics is where computer vision delivers its highest operational value — counting people, tracking objects across camera zones, detecting specific activities, and alerting on defined events — but requires careful architecture design for real-time processing performance that frame-by-frame detection-only approaches cannot achieve.

The system covers object detection with your specified classes, DeepSORT or ByteTrack multi-object tracking, zone-based counting and event triggering, activity recognition for defined activities, real-time stream processing or batch video analysis, and an alert system with your notification channel integration.