Object Tracking Evaluation: BoT-SORT & ByteTrack with YOLOv8

In recent years, rapid advancements in computer vision have led to the integration of innovative use cases into our daily lives spanning from surveillance to autonomous vehicles being out on our streets . Notably, object detection and tracking have witnessed the emergence of powerful models such as You Only Look Once (YOLO), renowned for its real-time efficiency and accuracy. In this project, we conducted a comparative analysis of two tracking algorithms embedded within YOLOv8, BoT-SORT and ByteTrack, aiming to benchmark their performance in terms of Multiple Object Tracking Accuracy (MOTA), Multiple Object Tracking Precision (MOTP), and Frames Per Second (FPS). YOLOv8, the latest iteration in the YOLO family developed by Ultralytics, was trained on a diverse dataset encompassing vehicles of varying shapes, sizes, and colors. Our benchmarks reveal that BoT-SORT exhibits greater accuracy and precision, whereas ByteTrack outperforms in terms of speed and computational efficiency. Consequently, the choice between BoT-SORT and ByteTrack hinges on the specific priority of either accuracy or speed.