boxmot: a pluggable multi-object tracking framework with unified Python and C++ implementations

What it solves

BoxMOT provides a unified interface for multi-object tracking (MOT) workflows. It eliminates the need to rebuild the entire detector and tracker stack for every experiment by offering a pluggable system where trackers, detectors, and ReID (re-identification) modules can be easily swapped.

How it works

The project provides a CLI and a Python API that allows users to run tracking workflows. It supports both axis-aligned bounding boxes (AABB) and oriented bounding boxes (OBB). The system can be able to integrate with various detectors (like YOLO) and ReID models. For production environments, it offers native C++ implementations of trackers that maintain the same metrics as the Python versions.

Who it’s for

It is designed for researchers and developers working on computer vision, specifically those implementing multi-object tracking systems, conducting benchmark evaluations, or deploying tracking software into production C++ applications.

Highlights

• Unified Interface: Single CLI/API for tracking, generation, evaluation, tuning, and ReID training.
• Swappable Components: Easily switch between different trackers (e.g., ByteTrack, BoTSORT, OccluBoost) while keeping the same plumbing.
• Production Ready: Optional native C++ tracker implementations for high-performance deployment.
• Comprehensive Benchmarking: Built-in support for cached benchmark evaluation and research loops.
• Flexible Bounding Boxes: Supports both AABB and OBB tracking paths.