BoxMOT: Pluggable python and c++ SOTA multi-object tracking modules with support for axis-aligned and oriented bounding boxes
pip install boxmot 
Pluggable Python and C++ multi-object tracking modules for axis-aligned and oriented bounding box detections from any model.
Docs • Installation • Modes • API Reference • Trackers • Contributing
BoxMOT gives you one CLI and one Python API for running modern multi-object tracking workflows. It covers direct tracking, cached benchmark evaluation, tuning, research loops, ReID training and evaluation, and ReID export without forcing you to rebuild the detector and tracker stack for each experiment.
Why BoxMOT
- One interface for
track,generate,eval,tune,research,train,eval-reid, andexport. - Swappable trackers with shared detector and ReID plumbing.
- Benchmark-oriented workflows with reusable detections and embeddings.
- Support for both AABB and OBB tracking paths.
- Optional production-ready native C++ tracker implementations with the same metrics as the Python path, opted into via
--tracker-backend cppand embeddable in standalone C++ projects via CMake (see Native C++ Integration). - Public Python API for embedding the same workflows in applications and notebooks.
Installation
BoxMOT supports Python 3.10 through 3.13.
pip install boxmot
boxmot --help
For mode-specific extras such as yolo, evolve, research, onnx, openvino, and tflite, see the installation guide.
Benchmark Results
| Tracker | Status | MOT17 ablation | SportsMOT val | MMOT OBB test | OBB | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| HOTA | MOTA | IDF1 | HOTA | MOTA | IDF1 | HOTA | MOTA | IDF1 | |||
| occluboost | ✅ | 70.47 (70.48) |
78.32 (78.31) |
84.14 (84.14) |
83.17 | 97.48 | 89.36 | 28.14 | 28.21 | 29.66 | ✅ |
| botsort | ✅ | 69.44 (69.43) |
78.24 (78.26) |
81.94 (82.00) |
76.93 | 98.11 | 78.30 | 52.27 | 45.45 | 61.36 | ✅ |
| boosttrack | ✅ | 69.25 (—) |
75.91 (—) |
83.20 (—) |
76.32 | 97.08 | 77.82 | 42.88 | 32.80 | 48.44 | ✅ |
| strongsort | ✅ | 68.05 (—) |
76.19 (—) |
80.76 (—) |
79.80 | 97.31 | 80.27 | 49.75 | 43.64 | 57.39 | ✅ |
| deepocsort | ✅ | 67.95 (—) |
75.83 (—) |
80.54 (—) |
79.51 | 97.94 | 79.59 | 50.43 | 43.93 | 58.39 | ✅ |
| bytetrack | ✅ | 67.68 (67.75) |
78.04 (78.03) |
79.16 (79.38) |
67.93 | 97.25 | 76.90 | 33.97 | 33.72 | 39.74 | ✅ |
| hybridsort | ✅ | 67.31 (—) |
74.09 (—) |
78.87 (—) |
81.14 | 98.07 | 81.88 | 53.86 | 44.63 | 63.22 | ✅ |
| ocsort | ✅ | 66.44 (66.44) |
74.55 (74.55) |
77.90 (77.90) |
76.34 | 96.60 | 75.64 | 28.57 | 26.19 | 29.95 | ✅ |
| sfsort | ✅ | 62.65 (62.66) |
76.87 (76.74) |
69.18 (69.18) |
75.73 | 98.39 | 72.99 | 44.19 | 44.27 | 46.25 | ✅ |
MMOT OBB results report the TrackEval Class Avg (Cls) across all eight categories on the 5,466-frame test split, using yolo11l_3ch detections and lmbn_n_duke ReID features. All trackers use their Python backend.
Py (C++); — unavailable. See Benchmark Workflows.
Related guides:
Minimal Usage
CLI:
boxmot track --detector yolo26n --reid lmbn_n_duke --tracker occluboost --source 0 --save --show
Python:
import numpy as np
from boxmot.trackers import OccluBoost
tracker = OccluBoost()
# dets: (N, 6) array with [x1, y1, x2, y2, conf, cls] per detection
dets = np.array([[100, 200, 300, 400, 0.9, 0]], dtype=np.float32)
img = np.zeros((480, 640, 3), dtype=np.uint8) # current frame
# tracks: (M, 8) array with [x1, y1, x2, y2, id, conf, cls, det_ind] per track
tracks = tracker.update(dets, img)
print(tracks)
Contributing
Start with CONTRIBUTING.md and the contributor docs.
Contributors
Support and Citation
- Bugs and feature requests: GitHub Issues
- Questions and discussion: GitHub Discussions or Discord
- Citation metadata: CITATION.cff
- Commercial support:
box-mot@outlook.com