open-mmlab/mmaction
An open-source PyTorch toolbox for video action understanding, implementing deep learning models for action recognition, temporal action detection, and spatial-temporal action localization.

MMAction is a computer vision toolbox built on PyTorch for understanding human actions in videos. It provides implementations of popular deep learning architectures including TSN, I3D, SlowFast, R(2+1)D, CSN for action recognition, and SSN for temporal action detection. The toolbox supports multiple benchmark datasets and follows a modular design with reusable components like backbones and sampling schemes.
Frequently asked
- What is open-mmlab/mmaction?
- An open-source PyTorch toolbox for video action understanding, implementing deep learning models for action recognition, temporal action detection, and spatial-temporal action localization.
- Is mmaction open source?
- Yes — open-mmlab/mmaction is open source, released under the Apache-2.0 license.
- What language is mmaction written in?
- open-mmlab/mmaction is primarily written in Python.
- How popular is mmaction?
- open-mmlab/mmaction has 1.9k stars on GitHub.
- Where can I find mmaction?
- open-mmlab/mmaction is on GitHub at https://github.com/open-mmlab/mmaction.