Résumé
The goal of human tracking is to detect people in ascene and assign them a unique identifier that the tracker willfollow across multiple frames. Our tracker, FORT, implementsdeep learning solutions such as, YOLOv7 for detection, ResNeXt-50 for feature extraction and re-identification and an adaptedKalman filter for tracking. The goal is to present a real timetracking solution for the complex environment of top-view, fisheyeimages. The proposed solution is then compared with the BoTSORTand StrongSORT trackers on a custom fisheye MultipleObject Tracking (MOT) Challenge dataset