New AI tracking algorithms package (github.com)

A new package called "Trackers" has been launched, offering modular re-implementations of leading multi-object tracking algorithms, including SORT and ByteTrack, with support for OC-SORT coming soon. Released under the Apache 2.0 license, this package allows developers to seamlessly integrate any detection model with various tracking algorithms, enhancing flexibility and functionality for AI applications. Installation is straightforward for Python environments starting from version 3.10, and users can also delve into unreleased features by installing from source. This development is significant for the AI and machine learning community as it streamlines the integration of different algorithms, encouraging collaboration and experimentation. The modular design facilitates the use of diverse object detection models, thus expanding the opportunities for innovation in areas such as video analysis and autonomous systems. The performance metrics, showcased through benchmarks, reveal that ByteTrack outperforms SORT in various evaluation criteria, including MOT17 HOTA, IDF1, and MOTA, which highlights the potency of these tracking algorithms in real-world scenarios. By fostering open collaboration, Trackers aims to enhance the efficiency and effectiveness of multi-object tracking in computer vision projects.