ByteTrack<T>

Models & Types · AiDotNet.ComputerVision.Tracking

ByteTrack: Multi-Object Tracking by Associating Every Detection Box.

Inheritance: ObjectTrackerBase<T> → ByteTrack<T>

For Beginners

ByteTrack improves upon SORT by using almost all detection boxes (including low-confidence ones) for tracking. It uses a two-stage association: first matching high-confidence detections, then low-confidence ones.

How It Works

Key features:

  • Associates all detection boxes (high and low confidence)
  • Two-stage matching strategy
  • Recovers occluded objects using low-score detections
  • State-of-the-art performance on MOT benchmarks

This follows the official BYTETracker: an xyah Kalman filter per track, IoU distance with detection scores fused in, lapjv-style assignment that accepts a pair only when its cost is at most the threshold (0.8, 0.5 and 0.7 for the high-score, low-score and unconfirmed rounds), and new tracks only from detections scoring at least track_thresh + 0.1.

Reference: Zhang et al., “ByteTrack: Multi-Object Tracking by Associating Every Detection Box”, ECCV 2022

Constructors

Constructor Summary
ByteTrack(TrackingOptions<T> options) Creates a new ByteTrack tracker.

Properties

Property Summary
string Name { get; }

Methods

Method Summary
TrackingResult<T> Update(List<Detection<T>> detections)
void Reset()