JonathonLuiten / TrackEval

HOTA (and other) evaluation metrics for Multi-Object Tracking (MOT).
MIT License
976 stars 243 forks source link

🐛 HOTA: fix numpy error when either ground truth or tracks are empty. #107

Open OlivierLDff opened 1 year ago

OlivierLDff commented 1 year ago

fixes a bug where a numpy error occurred when one of the shape dimensions of the data was 0. This was causing the potential matches between IDs in each timestep to not be counted correctly. The fix introduced by if len(gt_ids_t) and len(tracker_ids_t): ensures that the potential matches are only counted if both gt_ids_t and tracker_ids_t are non-empty, which prevents the numpy error from occurring. This ensures that the potential matches are calculated correctly, resulting in more accurate tracking results.