CLI
BoxMOT exposes one command group for all supported workflows:
Core idea
MODEselects the workflow such astrack,generate, oreval.--detectorselects the detector backend or profile.--reidselects the appearance model or profile.--trackerselects the tracker implementation and its YAML config.
Legacy aliases such as --yolo-model, --reid-model, and --tracking-method are not part of the current CLI.
Common examples
Track a video:
boxmot track --detector yolov8n --reid osnet_x0_25_msmt17 --tracker botsort --source video.mp4 --save
Evaluate a tracker on a benchmark:
Export a ReID model:
Run GEPA-based research:
boxmot research --benchmark mot17-ablation --tracker bytetrack --proposal-model openai/gpt-5.4 --max-metric-calls 24
Direct source vs benchmark configs
Use track when you already have a concrete source such as 0, video.mp4, path/, or rtsp://....
Use benchmark-driven modes when you want BoxMOT to resolve dataset, detector, and ReID profiles automatically from config files:
Full argument tables
Each mode page includes its own generated CLI argument table. Direct links: