Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/MCG-NJU/VideoMAE
[NeurIPS 2022 Spotlight] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
https://github.com/MCG-NJU/VideoMAE
action-recognition mae masked-autoencoder neurips-2022 pytorch self-supervised-learning transformer video-analysis video-representation-learning video-transformer video-understanding vision-transformer
Last synced: about 2 months ago
JSON representation
[NeurIPS 2022 Spotlight] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
- Host: GitHub
- URL: https://github.com/MCG-NJU/VideoMAE
- Owner: MCG-NJU
- License: other
- Created: 2022-03-23T06:15:50.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2023-12-08T13:44:48.000Z (about 1 year ago)
- Last Synced: 2024-08-01T02:25:18.109Z (4 months ago)
- Topics: action-recognition, mae, masked-autoencoder, neurips-2022, pytorch, self-supervised-learning, transformer, video-analysis, video-representation-learning, video-transformer, video-understanding, vision-transformer
- Language: Python
- Homepage: https://arxiv.org/abs/2203.12602
- Size: 547 KB
- Stars: 1,283
- Watchers: 16
- Forks: 128
- Open Issues: 40
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- Awesome-Reasoning-Foundation-Models - [Code
- StarryDivineSky - MCG-NJU/VideoMAE - 95%) 和管掩蔽策略，为自监督视频预训练创建具有挑战性的任务。VideoMAE 可以作为未来自监督视频预训练研究的简单但强大的基线。适用于不同规模的视频数据集，在 Kinects-400 上可以达到 87.4%（Kinetics-400是一个大规模，高质量的YouTube视频网址数据集，其中包含各种以人为本的行动。该数据集包含 400 个人类动作类，每个动作至少有 400 个视频剪辑。每个剪辑持续大约 10 秒，并且取自不同的 YouTube 视频。这些动作以人类为中心，涵盖广泛的类别，包括演奏乐器等人与物体的交互，以及握手等人与人的交互。），在 Something-Something V2 （大型的带有标签的记录了人类与日常生活中的一些物体之间的动作数据集）上可以达到 75.4%，在 UCF101 上可以达到 91.3%（UCF-101（2012）包含13,320个视频（共27个小时），101个人类行为类别，如运动、乐器和人物交互等。），在 HMDB51（HMDB51包含51类动作，共有6849个视频，每个动作至少包含51个视频，分辨率320*240,。来自于YouTube，google视频等，共２G） 上可以达到 62.6%。 (其他_机器视觉 / 网络服务_其他)
README
# Official PyTorch Implementation of VideoMAE (NeurIPS 2022 Spotlight).
[](https://creativecommons.org/licenses/by-nc/4.0/)
[](https://huggingface.co/models?other=videomae)[](https://huggingface.co/spaces/sayakpaul/video-classification-ucf101-subset)[](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/video_classification.ipynb)
[](https://paperswithcode.com/sota/action-recognition-in-videos-on-something?p=videomae-masked-autoencoders-are-data-1)
[](https://paperswithcode.com/sota/action-classification-on-kinetics-400?p=videomae-masked-autoencoders-are-data-1)
[](https://paperswithcode.com/sota/action-recognition-on-ava-v2-2?p=videomae-masked-autoencoders-are-data-1)
[](https://paperswithcode.com/sota/self-supervised-action-recognition-on-ucf101?p=videomae-masked-autoencoders-are-data-1)
[](https://paperswithcode.com/sota/self-supervised-action-recognition-on-hmdb51?p=videomae-masked-autoencoders-are-data-1)
> [**VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training**](https://arxiv.org/abs/2203.12602)
> [Zhan Tong](https://github.com/yztongzhan), [Yibing Song](https://ybsong00.github.io/), [Jue Wang](https://juewang725.github.io/), [Limin Wang](http://wanglimin.github.io/)
Nanjing University, Tencent AI Lab
## 📰 News
**[2023.4.18]** 🎈Everyone can download **Kinetics-400**, which is used in VideoMAE, from [this link](https://opendatalab.com/Kinetics-400).
**[2023.4.18]** Code and pre-trained models of [VideoMAE V2](https://arxiv.org/abs/2303.16727) have been released! Check and enjoy this [repo](https://github.com/OpenGVLab/VideoMAEv2)!
**[2023.4.17]** We propose **[EVAD](https://arxiv.org/abs/2304.08451)**, an **end-to-end Video Action Detection** framework.
**[2023.2.28]** Our [VideoMAE V2](https://arxiv.org/abs/2303.16727) is accepted by **CVPR 2023**! 🎉
**[2023.1.16]** Code and pre-trained models for **Action Detection** in VideoMAE are [available](https://github.com/MCG-NJU/VideoMAE-Action-Detection)!
**[2022.12.27]** 🎈Everyone can download extracted **VideoMAE** features of **THUMOS**, **ActivityNet**, **HACS** and **FineAction** from [InternVideo](https://github.com/OpenGVLab/InternVideo/tree/main/Downstream/Temporal-Action-Localization#to-reproduce-our-results-of-internvideo).
**[2022.11.20]** 👀 VideoMAE is integrated into [](https://huggingface.co/spaces/sayakpaul/video-classification-ucf101-subset) and [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/video_classification.ipynb), supported by [@Sayak Paul](https://github.com/sayakpaul).
**[2022.10.25]** 👀 VideoMAE is integrated into [MMAction2](https://github.com/open-mmlab/mmaction2/tree/dev-1.x/configs/recognition/videomae), the results on Kinetics-400 can be reproduced successfully.
**[2022.10.20]** The pre-trained models and scripts of **ViT-S** and **ViT-H** are available!
**[2022.10.19]** The pre-trained models and scripts on **UCF101** are [available](MODEL_ZOO.md#UCF101)!
**[2022.9.15]** VideoMAE is accepted by **NeurIPS 2022** as a **spotlight** presentation! 🎉
**[2022.8.8]** 👀 VideoMAE is integrated into **official** [🤗HuggingFace Transformers](https://huggingface.co/docs/transformers/main/en/model_doc/videomae) now! [](https://huggingface.co/models?other=videomae)
**[2022.7.7]** We have updated new results on downstream AVA 2.2 benchmark. Please refer to our [paper](https://arxiv.org/abs/2203.12602) for details.
**[2022.4.24]** Code and pre-trained models are available now!
**[2022.3.24]** ~~Code and pre-trained models will be released here.~~ Welcome to **watch** this repository for the latest updates.
## ✨ Highlights
### 🔥 Masked Video Modeling for Video Pre-Training
VideoMAE performs the task of masked video modeling for video pre-training. We propose the **extremely high** masking ratio (90%-95%) and **tube masking** strategy to create a challenging task for self-supervised video pre-training.
### ⚡️ A Simple, Efficient and Strong Baseline in SSVP
VideoMAE uses the simple masked autoencoder and **plain ViT** backbone to perform video self-supervised learning. Due to the extremely high masking ratio, the pre-training time of VideoMAE is **much shorter** than contrastive learning methods (**3.2x** speedup). VideoMAE can serve as **a simple but strong baseline** for future research in self-supervised video pre-training.
### 😮 High performance, but NO extra data required
VideoMAE works well for video datasets of different scales and can achieve **87.4%** on Kinects-400, **75.4%** on Something-Something V2, **91.3%** on UCF101, and **62.6%** on HMDB51. To our best knowledge, VideoMAE is the **first** to achieve the state-of-the-art performance on these four popular benchmarks with the **vanilla ViT** backbones while **doesn't need** any extra data or pre-trained models.
## 🚀 Main Results
### ✨ Something-Something V2
| Method | Extra Data | Backbone | Resolution | #Frames x Clips x Crops | Top-1 | Top-5 |
| :------: | :--------: | :------: | :--------: | :---------------------: | :---: | :---: |
| VideoMAE | ***no*** | ViT-S | 224x224 | 16x2x3 | 66.8 | 90.3 |
| VideoMAE | ***no*** | ViT-B | 224x224 | 16x2x3 | 70.8 | 92.4 |
| VideoMAE | ***no*** | ViT-L | 224x224 | 16x2x3 | 74.3 | 94.6 |
| VideoMAE | ***no*** | ViT-L | 224x224 | 32x1x3 | 75.4 | 95.2 |
### ✨ Kinetics-400
| Method | Extra Data | Backbone | Resolution | #Frames x Clips x Crops | Top-1 | Top-5 |
| :------: | :--------: | :------: | :--------: | :---------------------: | :---: | :---: |
| VideoMAE | ***no*** | ViT-S | 224x224 | 16x5x3 | 79.0 | 93.8 |
| VideoMAE | ***no*** | ViT-B | 224x224 | 16x5x3 | 81.5 | 95.1 |
| VideoMAE | ***no*** | ViT-L | 224x224 | 16x5x3 | 85.2 | 96.8 |
| VideoMAE | ***no*** | ViT-H | 224x224 | 16x5x3 | 86.6 | 97.1 |
| VideoMAE | ***no*** | ViT-L | 320x320 | 32x4x3 | 86.1 | 97.3 |
| VideoMAE | ***no*** | ViT-H | 320x320 | 32x4x3 | 87.4 | 97.6 |
### ✨ AVA 2.2
Please check the code and checkpoints in [VideoMAE-Action-Detection](https://github.com/MCG-NJU/VideoMAE-Action-Detection).
| Method | Extra Data | Extra Label | Backbone | #Frame x Sample Rate | mAP |
| :------: | :----------: | :---------: | :------: | :------------------: | :--: |
| VideoMAE | Kinetics-400 | ✗ | ViT-S | 16x4 | 22.5 |
| VideoMAE | Kinetics-400 | ✓ | ViT-S | 16x4 | 28.4 |
| VideoMAE | Kinetics-400 | ✗ | ViT-B | 16x4 | 26.7 |
| VideoMAE | Kinetics-400 | ✓ | ViT-B | 16x4 | 31.8 |
| VideoMAE | Kinetics-400 | ✗ | ViT-L | 16x4 | 34.3 |
| VideoMAE | Kinetics-400 | ✓ | ViT-L | 16x4 | 37.0 |
| VideoMAE | Kinetics-400 | ✗ | ViT-H | 16x4 | 36.5 |
| VideoMAE | Kinetics-400 | ✓ | ViT-H | 16x4 | 39.5 |
| VideoMAE | Kinetics-700 | ✗ | ViT-L | 16x4 | 36.1 |
| VideoMAE | Kinetics-700 | ✓ | ViT-L | 16x4 | 39.3 |
### ✨ UCF101 & HMDB51
| Method | Extra Data | Backbone | UCF101 | HMDB51 |
| :------: | :----------: | :------: | :----: | :----: |
| VideoMAE | ***no*** | ViT-B | 91.3 | 62.6 |
| VideoMAE | Kinetics-400 | ViT-B | 96.1 | 73.3 |
## 🔨 Installation
Please follow the instructions in [INSTALL.md](INSTALL.md).
## ➡️ Data Preparation
Please follow the instructions in [DATASET.md](DATASET.md) for data preparation.
## 🔄 Pre-training
The pre-training instruction is in [PRETRAIN.md](PRETRAIN.md).
## ⤴️ Fine-tuning with pre-trained models
The fine-tuning instruction is in [FINETUNE.md](FINETUNE.md).
## 📍Model Zoo
We provide pre-trained and fine-tuned models in [MODEL_ZOO.md](MODEL_ZOO.md).
## 👀 Visualization
We provide the script for visualization in [`vis.sh`](vis.sh). Colab notebook for better visualization is coming soon.
## ☎️ Contact
Zhan Tong: [email protected]
## 👍 Acknowledgements
Thanks to [Ziteng Gao](https://sebgao.github.io/), Lei Chen, [Chongjian Ge](https://chongjiange.github.io/), and [Zhiyu Zhao](https://github.com/JerryFlymi) for their kind support.
This project is built upon [MAE-pytorch](https://github.com/pengzhiliang/MAE-pytorch) and [BEiT](https://github.com/microsoft/unilm/tree/master/beit). Thanks to the contributors of these great codebases.
## 🔒 License
The majority of this project is released under the CC-BY-NC 4.0 license as found in the [LICENSE](https://github.com/MCG-NJU/VideoMAE/blob/main/LICENSE) file. Portions of the project are available under separate license terms: [SlowFast](https://github.com/facebookresearch/SlowFast) and [pytorch-image-models](https://github.com/rwightman/pytorch-image-models) are licensed under the Apache 2.0 license. [BEiT](https://github.com/microsoft/unilm/tree/master/beit) is licensed under the MIT license.
## ✏️ Citation
If you think this project is helpful, please feel free to leave a star⭐️ and cite our paper:
```
@inproceedings{tong2022videomae,
title={Video{MAE}: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training},
author={Zhan Tong and Yibing Song and Jue Wang and Limin Wang},
booktitle={Advances in Neural Information Processing Systems},
year={2022}
}
@article{videomae,
title={VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training},
author={Tong, Zhan and Song, Yibing and Wang, Jue and Wang, Limin},
journal={arXiv preprint arXiv:2203.12602},
year={2022}
}
```