UniFormer

This repo is the official implementation of "Uniformer: Unified Transformer for Efficient Spatiotemporal Representation Learning". It currently includes code and models for the following tasks:

• Image Classification

• Video Classification

• Object Detection (code will be released soon)

• Semantic Segmentation (code will be released soon)

• Pose Estimation (code will be released soon)

Updates

01/13/2022

[Initial commits]:

1. Pretrained models on ImageNet-1K, Kinetics-400, Kinetics-600, Something-Something V1&V2

2. The supported code and models for image classification and video classification are provided.

Introduction

UniFormer (Unified transFormer) is introduce in arxiv, which effectively unifies 3D convolution and spatiotemporal self-attention in a concise transformer format. We adopt local MHRA in shallow layers to largely reduce computation burden and global MHRA in deep layers to learn global token relation.

UniFormer achieves strong performance on video classification. With only ImageNet-1K pretraining, our UniFormer achieves 82.9%/84.8% top-1 accuracy on Kinetics-400/Kinetics-600, while requiring 10x fewer GFLOPs than other comparable methods (e.g., 16.7x fewer GFLOPs than ViViT with JFT-300M pre-training). For Something-Something V1 and V2, our UniFormer achieves 60.9% and 71.2% top-1 accuracy respectively, which are new state-of-the-art performances.

Main results on ImageNet-1K

Please see image_classification for more details.

More models with large resolution and token labeling will be released soon.

Main results on Kinetics-400

Please see video_classification for more details.

* Since Kinetics-600 is too large to train (>1 month in single node with 8 A100 GPUs), we provide model trained in multi node (around 2 weeks with 32 V100 GPUs), but the result is lower due to the lack of tuning hyperparameters.

Main results on Something-Something

Please see video_classification for more details.

Main results on downstream tasks

We have conducted extensive experiments on downstream tasks and achieved comparable results with SOTA models.

Code and models will be released in two weeks.

Cite Uniformer

If you find this repository useful, please use the following BibTeX entry for citation.

@misc{li2022uniformer,
      title={Uniformer: Unified Transformer for Efficient Spatiotemporal Representation Learning}, 
      author={Kunchang Li and Yali Wang and Peng Gao and Guanglu Song and Yu Liu and Hongsheng Li and Yu Qiao},
      year={2022},
      eprint={2201.04676},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

License

This project is released under the MIT license. Please see the LICENSE file for more information.

Contributors and Contact Information

UniFormer is maintained by Kunchang Li.

For help or issues using UniFormer, please submit a GitHub issue.

For other communications related to UniFormer, please contact Kunchang Li ([email protected]).