KaiyangZhou/deep-person-reid

Health v2 · maintenance only

Adoption not part of the score

4900 stars · 1215 forks observed · 2026-08-28

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

Torchreid is a PyTorch library for deep-learning person re-identification, supporting both image and video reid with end-to-end training and evaluation. It includes state-of-the-art models like OSNet, pretrained model zoo access, multi-dataset training, and export to ONNX/OpenVINO/TFLite.

Use cases

• train a person re-identification model in PyTorch
• evaluate reid models on standard benchmarks
• download pretrained OSNet weights for person reid
• run cross-dataset person re-identification evaluation
• extract person appearance embeddings for tracking
• export a reid model to ONNX or OpenVINO

When to choose

• you need person re-identification for surveillance, tracking, or image retrieval
• you want reproducible standard reid research protocols and pretrained models
• you need both image and video reid with multi-GPU training

When to avoid

• you need general object detection or classification rather than person reid
• you are not working in the PyTorch ecosystem
• you need a production-ready turnkey person search system with no ML experience

Facets

library · maturity stable

deep-learning machine-learning image-processing computer-vision benchmarking computer-vision deep-learning machine-learning image-processing python cross-platform person-reidentification pytorch metric-learning re-ranking osnet model-zoo video-reid cross-domain gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="KaiyangZhou/deep-person-reid")