AutoMix

<div align="center"> <h2><a href="https://arxiv.org/abs/2103.13027">AutoMix: Unveiling the Power of Mixup for Stronger Classifiers</a></h2> (ECCV 2022 Oral)

Zicheng Liu^*,1,2, Siyuan Li^*,1,2, Di Wu^1,2, Zhiyuan Chen¹, Lirong Wu^1,2, Stan Z. Li^†,1

¹Westlake University, ²Zhejiang University

</div> <p align="center"> <a href="https://arxiv.org/abs/2103.13027" alt="arXiv"> <img src="https://img.shields.io/badge/arXiv-2210.13452-b31b1b.svg?style=flat" /></a> <a href="https://github.com/Westlake-AI/AutoMix/blob/main/LICENSE" alt="license"> <img src="https://img.shields.io/badge/license-Apache--2.0-%23B7A800" /></a> <!-- <a href="https://colab.research.google.com/github/Westlake-AI/MogaNet/blob/main/demo.ipynb" alt="Colab"> <img src="https://colab.research.google.com/assets/colab-badge.svg" /></a> --> <a href="https://zhuanlan.zhihu.com/p/550300558" alt="license"> <img src="https://img.shields.io/badge/zhihu-automix-blue" /></a> </p>

We propose a novel automatic mixup (AutoMix) framework, where the mixup policy is parameterized and serves the ultimate classification goal directly. Specifically, AutoMix reformulates the mixup classification into two sub-tasks (i.e., mixed sample generation and mixup classification) with corresponding sub-networks and solves them in a bi-level optimization framework. For the generation, a learnable lightweight mixup generator, Mix Block, is designed to generate mixed samples by modeling patch-wise relationships under the direct supervision of the corresponding mixed labels. To prevent the degradation and instability of bi-level optimization, we further introduce a momentum pipeline to train AutoMix in an end-to-end manner. Extensive experiments on nine image benchmarks prove the superiority of AutoMix compared with state-of-the-arts in various classification scenarios and downstream tasks.

<p align="center"> <img src="https://user-images.githubusercontent.com/44519745/174272662-19ce57ad-7b08-4e73-81b1-3bb81fee2fe5.png" width=100% height=100% class="center"> </p> <!-- <details> <summary>Table of Contents</summary> <ol> <li><a href="#catalog">Catalog</a></li> <li><a href="#image-classification">Image Classification</a></li> <li><a href="#license">License</a></li> <li><a href="#acknowledgement">Acknowledgement</a></li> <li><a href="#citation">Citation</a></li> </ol> </details> -->

Catalog

We plan to update this timm implementation of AutoMix in a few months. Please watch us for the latest release or use our OpenMixup implementations.

• [x] Image Classification Code with OpenMixup [code]
• [x] CIFAR-10/100 and Tiny-ImageNet Training and Validation Code with timm [code]
• [ ] ImageNet-1K Training and Validation Code [code]
• [ ] Image Classification on Google Colab and Notebook Demo

Installation

Please check INSTALL.md for installation instructions.

Small-scale Image Classification

Please refer to OpenMixup implementations of CIFAR-100 and Tiny-ImageNet.

ImageNet Classification

1. Training and Validation

See TRAINING.md for ImageNet-1K training and validation instructions, or refer to our OpenMixup implementations. We released pre-trained models on OpenMixup.

<!-- Here is a notebook [demo](demo.ipynb) of AutoMix which run the steps to perform inference for image classification and generate mixup samples. -->

2. ImageNet-1K Trained Models

Please refer to mixup_benchmarks in OpenMixup implementations for results and models.

<p align="right">(<a href="#top">back to top</a>)</p>

License

This project is released under the Apache 2.0 license.

Acknowledgement

Our implementation is mainly based on the following codebases. We gratefully thank the authors for their wonderful works.

• pytorch-image-models: PyTorch image models, scripts, pretrained weights.
• OpenMixup: CAIRI Supervised, Semi- and Self-Supervised Visual Representation Learning Toolbox and Benchmark.

Citation

If you find this repository helpful, please consider citing:

@InProceedings{liu2022automix,
      title={AutoMix: Unveiling the Power of Mixup for Stronger Classifiers},
      author={Zicheng Liu and Siyuan Li and Di Wu and Zhiyuan Chen and Lirong Wu and Jianzhu Guo and Stan Z. Li},
      booktitle={European Conference on Computer Vision},
      pages={441--458},
      year={2022},
}

<p align="right">(<a href="#top">back to top</a>)</p>