Skip to content

Results and Artifacts

BoxMOT writes run outputs under the configured project/name directory.

Common artifacts

Depending on workflow and flags, you may see:

  • rendered video outputs
  • MOT-style .txt tracking files
  • cropped detections
  • benchmark summaries
  • tuning trial artifacts
  • GEPA state and candidate directories
  • exported ReID model files

Tracking outputs

track can write:

  • video or image outputs with overlays
  • MOT text files via --save-txt
  • crops via --save-crop

The public Python API also exposes structured result objects such as TrackRunResult, Results, and Tracks. See Results Objects.

Benchmark outputs

generate, eval, tune, and research also create reusable caches and evaluation artifacts tied to the selected benchmark.

Reproducibility tip

For repeated experiments, keep the same benchmark, detector, ReID, and tracker selections together under one project tree so cached detections and embeddings are easy to reuse.