ByteTrack: Multi-Object Tracking by Associating Every Detection Box

A clean implementation of ByteTrack for multi-object tracking using Kalman filtering and two-stage association.

Click to see higher quality video

output.mp4

Installation

pip install -r requirements.txt

Usage

import numpy as np
from bytetrack import BYTETracker

tracker = BYTETracker(track_thresh=0.5, track_buffer=30, match_thresh=0.8, low_thresh=0.1)

# detections: numpy array of shape (N, 5) -> [x1, y1, x2, y2, score]
detections = np.array([[100, 100, 200, 200, 0.9], [300, 300, 400, 400, 0.8]])

# tracks: numpy array of shape (M, 5) -> [x1, y1, x2, y2, track_id]
tracks = tracker.update(detections)

Demo

Face tracking demo using UniFace for detection:

python demo.py --source 0  # webcam
python demo.py --source video.mp4 --save output.mp4

References

• ByteTrack: ifzhang/ByteTrack - Original implementation
• UniFace: yakhyo/uniface - Face detection library