FoundYou: Real-Time Multi-Camera Person Re-Identification with Robust Gallery Matching and Active-Track Merging

We introduce FoundYou, a multi-camera person re-identification (ReID) system capable of real-time operation on commodity hardware with no camera calibration required. FoundYou combines YOLO26 detection, ByteTrack local tracking, and OSNet appearance embedding in a three-stage pipeline for resolving identities of people globally: (A) active tracks association through robust gallery distance, (B) archived track re-acquisition using FAISS acceleration, and (C) novel tracks generation. The subsequent post-hoc merge pass fixes early fragmentation before the galleries are mature enough. Face detection gives veto and fused score signal without being the primary detector. Runtime identity consistency validation catches swaps, duplicates, and unstable galleries. On synthetic multi-camera data, FoundYou achieves MOTA > 0.93 and IDF1 > 0.89 under standard MOT17 protocol at 18–24 fps on a laptop GPU.