xenonyu/mmaction2

OpenMMLab's Next Generation Video Understanding Toolbox and Benchmark

Introduction

English | 简体中文

MMAction2 is an open-source toolbox for video understanding based on PyTorch. It is a part of the OpenMMLab project.

The master branch works with PyTorch 1.3+.

Action Recognition Results on Kinetics-400

Spatio-Temporal Action Detection Results on AVA-2.1

Major Features

• Modular design

  We decompose the video understanding framework into different components and one can easily construct a customized video understanding framework by combining different modules.

• Support for various datasets

  The toolbox directly supports multiple datasets, UCF101, Kinetics-[400/600/700], Something-Something V1&V2, Moments in Time, Multi-Moments in Time, THUMOS14, etc.

• Support for multiple video understanding frameworks

  MMAction2 implements popular frameworks for video understanding:

  • For action recognition, various algorithms are implemented, including TSN, TSM, TIN, R(2+1)D, I3D, SlowOnly, SlowFast, CSN, Non-local, etc.

  • For temporal action localization, we implement BSN, BMN, SSN.

• Well tested and documented

  We provide detailed documentation and API reference, as well as unittests.

Changelog

v0.11.0 was released in 31/01/2021. Please refer to changelog.md for details and release history.

Benchmark

Model	input	io backend	batch size x gpus	MMAction2 (s/iter)	MMAction (s/iter)	Temporal-Shift-Module (s/iter)	PySlowFast (s/iter)
TSN	256p rawframes	Memcached	32x8	0.32	0.38	0.42	x
TSN	256p dense-encoded video	Disk	32x8	0.61	x	x	TODO
I3D heavy	256p videos	Disk	8x8	0.34	x	x	0.44
I3D	256p rawframes	Memcached	8x8	0.43	0.56	x	x
TSM	256p rawframes	Memcached	8x8	0.31	x	0.41	x
Slowonly	256p videos	Disk	8x8	0.32	TODO	x	0.34
Slowfast	256p videos	Disk	8x8	0.69	x	x	1.04
R(2+1)D	256p videos	Disk	8x8	0.45	x	x	x

Details can be found in benchmark.

ModelZoo

Supported methods for Action Recognition:

(click to collapse)

• TSN (ECCV'2016)
• TSM (ICCV'2019)
• TSM Non-Local (ICCV'2019)
• R(2+1)D (CVPR'2018)
• I3D (CVPR'2017)
• I3D Non-Local (CVPR'2018)
• SlowOnly (ICCV'2019)
• SlowFast (ICCV'2019)
• CSN (ICCV'2019)
• TIN (AAAI'2020)
• TPN (CVPR'2020)
• C3D (CVPR'2014)
• X3D (CVPR'2020)
• OmniSource (ECCV'2020)
• MultiModality: Audio (ArXiv'2020)

Supported methods for Temporal Action Detection:

(click to collapse)

• BSN (ECCV'2018)
• BMN (ICCV'2019)
• SSN (ICCV'2017)

Supported methods for Spatial Temporal Action Detection:

(click to collapse)

• SlowOnly+Fast R-CNN (ICCV'2019)
• SlowFast+Fast R-CNN (ICCV'2019)

Results and models are available in the README.md of each method's config directory. A summary can be found in the model zoo page.

Dataset

Supported datasets:

Supported datasets for Action Recognition:

(click to collapse)

• UCF101 [ Homepage ] (CRCV-IR-12-01)
• HMDB51 [ Homepage ] (ICCV'2011)
• Kinetics-[400/600/700] [ Homepage ] (CVPR'2017)
• Something-Something V1 [ Homepage ] (ICCV'2017)
• Something-Something V2 [ Homepage ] (ICCV'2017)
• Moments in Time [ Homepage ] (TPAMI'2019)
• Multi-Moments in Time [ Homepage ] (ArXiv'2019)
• HVU [ Homepage ] (ECCV'2020)
• Jester [ Homepage ] (ICCV'2019)
• GYM [ Homepage ] (CVPR'2020)
• ActivityNet [ Homepage ] (CVPR'2015)

Supported datasets for Temporal Action Detection

(click to collapse)

• ActivityNet [ Homepage ] (CVPR'2015)
• THUMOS14 [ Homepage ] (THUMOS Challenge 2014)

Supported datasets for Spatial Temporal Action Detection

(click to collapse)

• AVA [ Homepage ] (CVPR'2018)
• UCF101-24 [ Homepage ] (CRCV-IR-12-01)
• JHMDB [ Homepage ] (ICCV'2013)

Installation

Please refer to install.md for installation.

Data Preparation

Please refer to data_preparation.md for a general knowledge of data preparation. The supported datasets are listed in supported_datasets.md

Get Started

Please see getting_started.md for the basic usage of MMAction2. There are also tutorials for learn about configs, finetuning models, adding new dataset, designing data pipeline, adding new modules, exporting model to onnx and customizing runtime settings.

A Colab tutorial is also provided. You may preview the notebook here or directly run on Colab.

FAQ

Please refer to FAQ for frequently asked questions.

License

This project is released under the Apache 2.0 license.

Citation

If you find this project useful in your research, please consider cite:

@misc{2020mmaction2,
    title={OpenMMLab's Next Generation Video Understanding Toolbox and Benchmark},
    author={MMAction2 Contributors},
    howpublished = {\url{https://github.com/open-mmlab/mmaction2}},
    year={2020}
}

Contributing

We appreciate all contributions to improve MMAction2. Please refer to CONTRIBUTING.md in MMCV for more details about the contributing guideline.

Acknowledgement

MMAction2 is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors who implement their methods or add new features, as well as users who give valuable feedbacks. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new models.

Projects in OpenMMLab

• MMCV: OpenMMLab foundational library for computer vision.
• MMClassification: OpenMMLab image classification toolbox and benchmark.
• MMDetection: OpenMMLab detection toolbox and benchmark.
• MMDetection3D: OpenMMLab's next-generation platform for general 3D object detection.
• MMSegmentation: OpenMMLab semantic segmentation toolbox and benchmark.
• MMAction2: OpenMMLab's next-generation video understanding toolbox and benchmark.
• MMTracking: OpenMMLab video perception toolbox and benchmark.
• MMPose: OpenMMLab pose estimation toolbox and benchmark.
• MMEditing: OpenMMLab image and video editing toolbox.