Benchmark Workflows
BoxMOT's benchmark-driven workflows use a layered config system rather than forcing you to pass every path manually.
Resolution model
When you run:
BoxMOT resolves:
- the benchmark YAML under
boxmot/configs/benchmarks - the associated dataset YAML under
boxmot/configs/datasets - the associated detector profile under
boxmot/configs/detectors - the associated ReID profile under
boxmot/configs/reid - the tracker YAML under
boxmot/configs/trackers
Typical pipeline
boxmot generate --benchmark mot17-ablation
boxmot eval --benchmark mot17-ablation --tracker bytetrack
boxmot tune --benchmark mot17-ablation --tracker bytetrack
boxmot research --benchmark mot17-ablation --tracker bytetrack --proposal-model openai/gpt-5.4
Overrides
You can override detector and ReID selections explicitly:
boxmot eval \
--benchmark mot17-ablation \
--detector yolo11s_obb \
--reid lmbn_n_duke \
--tracker boosttrack