CrowdTrack-MOT

Pedestrian multi-object-tracking benchmark in MOTChallenge format, derived from Loseevaya/CrowdTrack. 33 CCTV / surveillance sequences, single class person, ~702k GT boxes, 1920×1080.

What changed vs. the original

1. Converted to MOT format and added a seqinfo.ini per sequence. The original ships two heterogeneous annotation formats (a few sequences as plain gt.txt, the rest as per-frame LabelMe JSON) — both are unified into a single MOT gt.txt. (frameRate=25 in seqinfo.ini is a hardcoded placeholder — the original FPS is unknown.)
2. Boxes clipped to the frame. Some original boxes extended past the image border (and a few were stored with inverted/negative width-height); these are normalized to top-left + positive size and clipped to the frame bounds.

Layout

train/
  trackNNNN/
    img1/000001.<ext> ...     # frames, 1-indexed MOT naming (.jpg or .png)
    gt/gt.txt                 # frame,id,x,y,w,h,1,1,1   (class: person)
    seqinfo.ini

Download

hf download Fleyderer/CrowdTrack-MOT --repo-type dataset --local-dir CrowdTrack-MOT

Works as a benchmark in BoxMOT:

boxmot eval --benchmark crowdtrack-mot --detector <det> --reid <reid> --tracker <trk>

Credits

• Original benchmark: CrowdTrack: A Benchmark for Difficult Multiple Pedestrian Tracking in Real Scenarios — Teng Fu, Yuwen Chen, Zhuofan Chen, Mengyang Zhao, Bin Li, Xiangyang Xue (arXiv:2507.02479).
• Source data: Loseevaya/CrowdTrack — all imagery and original annotations.
• MOT conversion / clipping / packaging: Fleyderer.

Citation

If you use this dataset, please cite the original CrowdTrack paper:

@article{fu2025crowdtrack,
  title   = {CrowdTrack: A Benchmark for Difficult Multiple Pedestrian Tracking in Real Scenarios},
  author  = {Fu, Teng and Chen, Yuwen and Chen, Zhuofan and Zhao, Mengyang and Li, Bin and Xue, Xiangyang},
  journal = {arXiv preprint arXiv:2507.02479},
  year    = {2025}
}