Zero-Shot Video Action Recognition for Retail Shoplifting Prevention
Spatio-temporal 3D CNN analyzing CCTV video streams to classify concealment and suspicious shelf-clearing behaviors in real time.
Project Overview
Processes multi-camera RTSP surveillance feeds. Employs a lightweight VideoMAE / SlowFast 3D convolutional model to detect anomalous shoplifting actions (pocketing merchandise, concealing items in bags) while ignoring normal customer browsing, raising instant store security alerts.
Processes multi-camera RTSP surveillance feeds. Employs a lightweight VideoMAE / SlowFast 3D convolutional model to detect anomalous shoplifting actions (pocketing merchandise, concealing items in bags) while ignoring normal customer browsing, raising instant store security alerts.
Spatio-temporal 3D CNN analyzing CCTV video streams to classify concealment and suspicious shelf-clearing behaviors in real time.