noahcao/OC_SORT

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1131 stars · 154 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

OC-SORT is a pure motion-model-based multi-object tracker for video, improving on SORT by fixing Kalman filter limitations to handle occlusion and non-linear motion. It is simple, online, and real-time, with a flexible design that integrates with different detectors and appearance-based matching modules.

Use cases

• track multiple objects in video in real time
• improve tracking robustness in crowded scenes
• handle non-linear object motion in tracking
• track objects through occlusions
• plug a tracker into different object detectors
• benchmark a tracker on MOT17, MOT20, or DanceTrack

When to choose

• you need a simple, online, real-time multi-object tracker without heavy appearance models
• your scenes involve occlusion or non-linear motion
• you want a well-cited CVPR 2023 tracker with Python and C++ support

When to avoid

• you need appearance-based re-identification as a first-class feature (consider Deep OC-SORT)
• you need a full tracking framework with training pipelines rather than a tracker
• you need a non-PyTorch deep learning stack

Facets

library · maturity stable

computer-vision machine-learning deep-learning computer-vision machine-learning deep-learning python cpp windows multi-object-tracking sort kalman-filter pytorch cvpr2023 video-tracking object-detection linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="noahcao/OC_SORT")