FoundationVision/ByteTrack
A PyTorch implementation of multi-object tracking in video using deep learning detection boxes.

Velocity · 7d
+3.7
★ / day
Trend
→steady
star history
ByteTrack is a computer vision research project for multi-object tracking (MOT) published at ECCV 2022. It associates every detection box, including low-confidence ones, to track objects across video frames. Built on PyTorch, it provides real-time tracking performance with models and a demo deployed on Huggingface Spaces via Gradio.