Zhongdao/Towards-Realtime-MOT

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2444 stars · 534 forks observed · 2026-08-28

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

A PyTorch codebase for the Joint Detection and Embedding (JDE) model, a fast multiple-object tracker that learns object detection and appearance embedding simultaneously in a shared neural network. It achieves MOTA 64%+ on MOT-16 at 22-38 FPS and includes training data, pretrained models, and evaluation tooling.

Use cases

• track multiple pedestrians in a video in real time
• run multi-object tracking on raw video footage
• train a joint detection and embedding tracker
• evaluate a tracker on the MOT-16 benchmark
• deploy a fast MOT system for surveillance video
• benchmark multi-object tracking algorithms

When to choose

• you need near-realtime multi-object tracking with detection included in the pipeline
• you want a research baseline with pretrained models and MOT-16 evaluation
• you need to train or extend a joint detection and embedding tracker in PyTorch

When to avoid

• you need actively maintained code with recent dependency support
• you want a pure C++ or production-ready deployment without Python
• you need tracking of general object classes beyond pedestrians
• you cannot use a GPU

Facets

library · maturity maintenance

computer-vision machine-learning deep-learning image-processing computer-vision machine-learning deep-learning python multi-object-tracking tracking-by-detection pedestrian-tracking pytorch eccv-2020 realtime object-detection appearance-embedding linux gpu docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Zhongdao/Towards-Realtime-MOT")