tajwarchy/multi-camera-people-tracking

This repository implements a multi-camera pedestrian tracking system: synchronized camera frames are processed into per-camera detections, local tracks and appearance embeddings; cross-camera association assigns global identities; calibrated projections and those identities drive a bird’s-eye view and composite video. Separate utilities prepare and score ReID evaluation crops. The sampled sources establish the core runtime calls; calibration and frame-extraction roles are documented in the README, so links to later stages are shown only where documented.