VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking
Limin Wang, Bingkun Huang, Zhiyu Zhao, Zhan Tong, Yinan He, Yi Wang, Yali Wang, and Yu Qiao
Nanjing University, Shanghai AI Lab, CAS
[2023.05.29] VideoMAE V2-g features for THUMOS14 and FineAction datasets are available at TAD.md now.
[2023.05.11] We have supported testing of our distilled models at MMAction2 (dev version)! See PR#2460.
[2023.05.11] The feature extraction script for TAD datasets has been released! See instructions at TAD.md.
[2023.04.19] ViT-giant model weights have been released! You can get the download links from MODEL_ZOO.md.
[2023.04.18] Code and the distilled models (vit-s & vit-b) have been released!
[2023.04.03] Code and models will be released soon.
We now provide the model weights in MODEL_ZOO.md. We have additionally provided distilled models in MODEL_ZOO.
Model | Dataset | Teacher Model | #Frame | K710 Top-1 | K400 Top-1 | K600 Top-1 |
---|---|---|---|---|---|---|
ViT-small | K710 | vit_g_hybrid_pt_1200e_k710_ft | 16x5x3 | 77.6 | 83.7 | 83.1 |
ViT-base | K710 | vit_g_hybrid_pt_1200e_k710_ft | 16x5x3 | 81.5 | 86.6 | 85.9 |
Please follow the instructions in INSTALL.md.
Please follow the instructions in DATASET.md for data preparation.
The pre-training instruction is in PRETRAIN.md.
The fine-tuning instruction is in FINETUNE.md.
If you find this repository useful, please use the following BibTeX entry for citation.
@InProceedings{wang2023videomaev2,
author = {Wang, Limin and Huang, Bingkun and Zhao, Zhiyu and Tong, Zhan and He, Yinan and Wang, Yi and Wang, Yali and Qiao, Yu},
title = {VideoMAE V2: Scaling Video Masked Autoencoders With Dual Masking},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {14549-14560}
}
@misc{videomaev2,
title={VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking},
author={Limin Wang and Bingkun Huang and Zhiyu Zhao and Zhan Tong and Yinan He and Yi Wang and Yali Wang and Yu Qiao},
year={2023},
eprint={2303.16727},
archivePrefix={arXiv},
primaryClass={cs.CV}
}