CCTV Vape & Anomaly Detection

A real-time computer-vision system that watches existing CCTV feeds for vaping and other behavioural anomalies, scores what it sees, and pushes a WhatsApp alert to whoever is on duty. It never relies on spotting the device — at camera distance that is a few pixels — so it fuses weak signals instead: pose-estimated wrist-to-mouth distance normalised in head-units, a temporal inhale-hold-exhale rhythm, and a vapour channel. Every alert states which signals fired, what they scored and what the threshold was, and says plainly that it flags rather than confirms.