cj-mills / byte-track-eigen

ByteTrack-Eigen is a C++ implementation of the ByteTrack object tracking method, leveraging the Eigen library for high-performance matrix and vector operations.

ByteTrack-Eigen

Introduction

ByteTrack-Eigen is a C++ implementation of the ByteTrack object tracking method, leveraging the Eigen library for high-performance matrix and vector operations. This library is designed for tracking objects in video frames using Kalman Filters and the Hungarian algorithm.

Demo Video

yolox-byte-track-demo.mp4

Source Video by RDNE Stock project from Pexels: (link)

Features

• Object tracking using Kalman Filters and Hungarian algorithm.
• Handling of occlusions and re-identifying lost objects.
• Utilization of the Eigen library for optimized matrix and vector operations.

Components

• BaseTrack: Foundational class for object tracking.
• BoundingBoxTrackUtils: Utility functions for track operations.
• BYTETracker: Main class for the BYTE tracking algorithm.
• HungarianAlgorithmEigen: Hungarian Algorithm implementation with Eigen.
• KalmanBBoxTrack: Kalman filter-based bounding box tracking.
• KalmanFilter: Kalman filter implementation for tracking bounding boxes in image space.
• LinearAssignment: Solves linear assignment problems using the Hungarian Algorithm.
• BoundingBoxIoUMatching: Utility functions for bounding box operations.

Requirements

• CMake 3.20 or higher.
• C++17 or higher.
• Eigen library (automatically downloaded by CMake).

Building the Library

1. Clone the repository:

   git clone https://github.com/cj-mills/byte-track-eigen.git

2. Navigate to the project directory:

   cd byte-track-eigen

3. Build the project using CMake:

   mkdir build && cd build

   cmake ..

Usage

Here is an example of how to use ByteTrack-Eigen in your project:

// Example usage of ByteTrack-Eigen
#include "BYTETracker.h"

int main() {
    float track_thresh = 0.23f;
    int track_buffer = 30;
    float match_thresh = 0.8;
    int frame_rate = 30;
    BYTETracker tracker(track_thresh, track_buffer, match_thresh, frame_rate);
}

Demo Projects

• yolox-bytetrack-onnx-demo: A Visual Studio project demonstrating how to perform object tracking across video frames with YOLOX, ONNX Runtime, and the ByteTrack-Eigen library.
• unity-bytetrack-plugin: A simple native plugin for the Unity game engine, built in Visual Studio, that leverages the ByteTrack-Eigen library to perform real-time object tracking.

License

This project is licensed under the MIT License.

Acknowledgments

Special thanks to the Eigen library team and authors of the ByteTrack algorithm.