Awesome
[CVPR 2024] Official Implementation of AMD
The following ranking results are produced by vit-b
:
Asymmetric Masked Distillation for Pre-Training Small Foundation Models<br> Zhiyu Zhao, Bingkun Huang, Sen Xing, Gangshan Wu, Yu Qiao, and Limin Wang<br> Nanjing University, Shanghai AI Lab<br>
News ๐ฐ
[2024.3.27] Code and models have been released!<br>
[2024.2.29] Code and models will be released in the following days.<br>
[2024.2.27] AMD is accpeted by CVPR2024! ๐๐๐<br>
Main Results ๐
โจ Something-Something V2
Method | Extra Data | Backbone | Resolution | #Frames x Clips x Crops | Top-1 | Top-5 |
---|---|---|---|---|---|---|
AMD | no | ViT-S | 224x224 | 16x2x3 | 70.2 | 92.5 |
AMD | no | ViT-B | 224x224 | 16x2x3 | 73.3 | 94.0 |
โจ Kinetics-400
Method | Extra Data | Backbone | Resolution | #Frames x Clips x Crops | Top-1 | Top-5 |
---|---|---|---|---|---|---|
AMD | no | ViT-S | 224x224 | 16x5x3 | 80.1 | 94.5 |
AMD | no | ViT-B | 224x224 | 16x5x3 | 82.2 | 95.3 |
โจ AVA 2.2
Method | Extra Data | Extra Label | Backbone | #Frame x Sample Rate | mAP |
---|---|---|---|---|---|
AMD | Kinetics-400 | โ | ViT-B | 16x4 | 29.9 |
AMD | Kinetics-400 | โ | ViT-B | 16x4 | 33.5 |
โจ UCF101 & HMDB51
Method | Extra Data | Backbone | UCF101 | HMDB51 |
---|---|---|---|---|
AMD | Kinetics-400 | ViT-B | 97.1 | 79.6 |
โจ ImageNet-1K
Method | Extra Data | Backbone | Resolution | Top-1 |
---|---|---|---|---|
AMD | no | ViT-S | 224x224 | 82.1 |
AMD | no | ViT-B | 224x224 | 84.6 |
Installation ๐จ
Please follow the instructions in INSTALL.md.
Data Preparation โก๏ธ
Please follow the instructions in DATASET.md for data preparation.
Pre-training ๐
The pre-training instruction is in PRETRAIN.md.
Fine-tuning โคด๏ธ
The fine-tuning instruction is in FINETUNE.md.
Model Zoo ๐
We provide pre-trained and fine-tuned models in MODEL_ZOO.md.
Acknowledgements ๐
This project is built upon VideoMAEv2 and MGMAE. Thanks to the contributors of these great codebases.
Citation โ๏ธ
If you find this repository useful, please use the following BibTeX entry for citation.
@misc{zhao2023amd,
title={Asymmetric Masked Distillation for Pre-Training Small Foundation Models},
author={Zhiyu Zhao and Bingkun Huang and Sen Xing and Gangshan Wu and Yu Qiao and Limin Wang},
year={2023},
eprint={2311.03149},
archivePrefix={arXiv},
primaryClass={cs.CV}
}