MMAction is an open source toolbox for action understanding based on PyTorch. It is a part of the open-mmlab project developed by Multimedia Laboratory, CUHK.
-
MMAction is capable of dealing with all of the tasks below.
- action recognition from trimmed videos
- temporal action detection (also known as action localization) in untrimmed videos
- spatial-temporal action detection in untrimmed videos.
-
Support for various datasets
Video datasets have emerging throughout the recent years and have greatly fostered the devlopment of this field. MMAction provides tools to deal with various datasets.
-
Support for multiple action understanding frameworks
MMAction implements popular frameworks for action understanding:
- For action recognition, various algorithms are implemented, including TSN, I3D, SlowFast, R(2+1)D, CSN.
- For temporal action detection, we implement SSN.
- For spatial temporal atomic action detection, a Fast-RCNN baseline is provided.
-
Modular design
The tasks in human action understanding share some common aspects such as backbones, and long-term and short-term sampling schemes. Also, tasks can benefit from each other. For example, a better backbone for action recognition will bring performance gain for action detection. Modular design enables us to view action understanding in a more integrated perspective.
The project is release under the Apache 2.0 license.
v0.1.0 (19/06/2019)
- MMAction is online!
v0.2.0 (15/03/2020)
- We build a diversified modelzoo for action recognition, which include popular algorithms (TSN, I3D, SlowFast, R(2+1)D, CSN). The performance is aligned with or better than the original papers.
Results and reference models are available in the model zoo.
Please refer to INSTALL.md for installation.
Update: for Docker installation, Please refer to DOCKER.md for using docker for this project.
Please refer to DATASET.md for a general knowledge of data preparation.
Detailed documents for the supported datasets are available in data_tools/.
Please refer to GETTING_STARTED.md for detailed examples and abstract usage.
We appreciate all contributions to improve MMAction. Please refer to CONTRUBUTING.md for the contributing guideline.
If you use our codebase or models in your research, please cite this work. We will release a technical report later.
@misc{mmaction2019,
author = {Yue Zhao, Yuanjun Xiong, Dahua Lin},
title = {MMAction},
howpublished = {\url{https://github.com/open-mmlab/mmaction}},
year = {2019}
}
If you have any question, please file an issue or contact the author:
Yue Zhao: thuzhaoyue@gmail.com