Skip to content

installation.md

Latest commit

 

History

History
87 lines (58 loc) · 3.39 KB

installation.md

File metadata and controls

87 lines (58 loc) · 3.39 KB
Raw

Installation

Requirements

BoxMOT supports Python 3.10 through 3.13.

Basic install

pip install boxmot
boxmot --help

This installs the core package. It is enough for simple tracking and Python API usage. Optional detector and ReID runtime packages are installed on first use when BoxMOT can resolve a safe package for the selected backend.

Mode-specific extras

BoxMOT keeps heavier workflow dependencies optional. Install the extras that match the modes and export targets you plan to use.

Workflow PyPI install Source checkout with uv Notes
track, generate, eval with common YOLO backends pip install "boxmot[yolo]" uv sync --extra yolo Preinstalls Ultralytics and YOLOX. If you skip this, BoxMOT can install some detector packages on first use.
train and eval-reid pip install boxmot uv sync Uses the built-in ReID training and evaluation stack. You still need to place the chosen ReID dataset under your --data-dir.
tune pip install "boxmot[evolve]" uv sync --extra evolve Installs Ray Tune, Optuna, Plotly, and related tuning dependencies.
research pip install "boxmot[research]" uv sync --extra research Installs GEPA for the code-evolution loop.
export --include onnx pip install "boxmot[onnx]" uv sync --extra onnx The default export path uses ONNX.
export --include openvino pip install "boxmot[openvino]" uv sync --extra openvino Usually paired with --include onnx.
export --include tflite pip install "boxmot[tflite]" uv sync --extra tflite Installs both TFLite export and LiteRT inference packages.

You can combine extras when needed:

uv sync --extra yolo --extra evolve --extra research
pip install "boxmot[yolo,evolve,research]"

When an optional ReID runtime is missing, BoxMOT attempts a first-use install with uv pip install when uv is available, otherwise with the active python -m pip. This covers ONNX Runtime, OpenVINO, LiteRT, and NVIDIA TensorRT. TensorRT still requires a compatible CUDA/NVIDIA stack for the installed wheel to import and run correctly.

Native C++ backends

Native C++ tracker backends are built lazily the first time you select --tracker-backend cpp. They are currently available for botsort, bytetrack, ocsort, occluboost, and sfsort.

Install the native build tools before using them:

  • C++17 compiler
  • CMake 3.16+
  • OpenCV 4.x
  • Eigen3 3.3+

Example:

boxmot track --detector yolov8n --tracker bytetrack --tracker-backend cpp --source video.mp4
boxmot eval --benchmark mot17 --split ablation --tracker bytetrack --tracker-backend cpp

The generated build files are kept under build/native/<tracker>/.

Verify the install

!!! example "Verify"

=== "CLI"

    Check the CLI:

    ```bash
    boxmot --help
    boxmot track --help
    ```

=== "Python"

    Smoke-test the Python API:

    ```python
    from boxmot import BoxMOT

    boxmot = BoxMOT(detector="yolov8n", reid="osnet_x0_25_msmt17", tracker="bytetrack")
    print(boxmot)
    ```

Next steps

  • Use Quickstart for a minimal path.
  • Use Modes Overview to decide between track, generate, eval, tune, research, train, eval-reid, and export.
  • Use the workflow table above to add the extras your workflow needs.