open-mmlab/mmflow

Health v2 · maintenance only

Adoption not part of the score

1062 stars · 116 forks observed · 2026-08-28

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

MMFlow is an open-source optical flow toolbox and benchmark built on PyTorch as part of the OpenMMLab project. It provides a unified, modular framework for implementing, training, and evaluating optical flow algorithms such as FlowNet, PWC-Net, and RAFT.

Use cases

• estimate optical flow between video frames
• train RAFT or PWC-Net on FlyingChairs
• benchmark optical flow models on Sintel and KITTI
• implement a custom optical flow architecture with modular components
• run pretrained optical flow inference in a computer vision pipeline

When to choose

• you need a PyTorch framework for optical flow estimation or evaluation
• you want pretrained optical flow models and standard benchmark datasets out of the box
• you prefer OpenMMLab-style modular configuration-driven training

When to avoid

• you need general video processing rather than flow estimation
• you require a framework actively adding new models, as development has slowed
• you work outside PyTorch or need lightweight production inference without GPU tooling

Facets

library · maturity maintenance

machine-learning computer-vision image-processing benchmarking computer-vision deep-learning machine-learning python windows optical-flow pytorch openmmlab model-zoo deep-learning-toolbox linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="open-mmlab/mmflow")