layumi/Person_reID_baseline_pytorch

Health v2 · maintenance only

Adoption not part of the score

4446 stars · 1028 forks observed · 2026-08-28

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

A small, friendly PyTorch baseline implementation for person and vehicle re-identification (ReID). It reproduces strong top-conference results with simple training code, supports memory-efficient bf16/fp16 training, and includes a beginner tutorial.

Use cases

• train a person re-identification model in pytorch
• re-identify people across camera views
• build a vehicle re-id baseline
• run image retrieval with re-ranking
• learn person reid with a tutorial
• benchmark on Market-1501 and MSMT17
• apply circle loss and random erasing tricks

When to choose

• you need a well-cited, easy-to-modify ReID baseline in PyTorch
• you have limited GPU memory (2GB with fp16)
• you want a tutorial-friendly entry into object re-identification
• you need standard benchmarks like Market-1501, MSMT17, CUHK-NP

When to avoid

• you need production-ready person tracking or detection pipelines
• you want a turnkey commercial surveillance product
• you work outside PyTorch or need non-vision retrieval tasks

Facets

library · maturity stable

machine-learning deep-learning image-processing computer-vision search-engine computer-vision machine-learning deep-learning image-processing python cross-platform person-reidentification vehicle-reid metric-learning re-ranking pytorch market-1501 circle-loss image-retrieval baseline tutorial gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="layumi/Person_reID_baseline_pytorch")