FoundationVision/ByteTrack

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

6654 stars · 1141 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

ByteTrack is a PyTorch-based multi-object tracking (MOT) library implementing the ECCV 2022 paper 'Multi-Object Tracking by Associating Every Detection Box'. It achieves state-of-the-art tracking accuracy (80.3 MOTA on MOT17) at real-time speeds by associating low-score detection boxes with tracklets to recover occluded objects.

Use cases

• track multiple objects in video with bounding boxes and identities
• track pedestrians in surveillance footage in real time
• recover occluded objects that low-score detectors miss
• improve IDF1 scores of existing trackers with a better association method
• run multi-object tracking at 30 FPS on a single GPU
• benchmark a tracker on MOT17 and MOT20 datasets

When to choose

• you need state-of-the-art multi-object tracking accuracy with real-time performance
• your detections include low-confidence boxes for occluded objects that other trackers discard
• you want a well-cited, research-backed tracker with pretrained models and deployment support
• you need a tracker that can be integrated on top of various state-of-the-art detectors

When to avoid

• you need simple single-object tracking rather than multi-object tracking
• you cannot run a GPU or need CPU-only real-time tracking
• you need a maintained project with frequent updates, as development has slowed since 2022
• you need out-of-the-box tracking without training or configuring a detector

Facets

library · maturity stable

computer-vision machine-learning image-processing computer-vision artificial-intelligence deep-learning python cross-platform multi-object-tracking object-detection pytorch real-time eccv-2022 video-analytics gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="FoundationVision/ByteTrack")