Home

Awesome

Reversible Column Networks

This repo is the official implementation of:

Reversible Column Networks

Yuxuan Cai, Yizhuang Zhou, Qi Han, Jianjian Sun, Xiangwen Kong, Jun Li, Xiangyu Zhang
MEGVII Technology
International Conference on Learning Representations (ICLR) 2023
[arxiv]

RevColV2: Exploring Disentangled Representations in Masked Image Modeling

Qi Han, Yuxuan Cai, Xiangyu Zhang
MEGVII Technology
[arxiv]

Updates

9/06/2023*
RevColv2 will be released soon!

3/15/2023*
RevCol Huge checkpoint for segmentation released! Add visualization tools.

3/9/2023*
Detection, Segmentation Code and Model Weights Released.

2/10/2023
RevCol model weights released.

1/21/2023
RevCol was accepted by ICLR 2023!

12/23/2022
Initial commits: codes for ImageNet-1k and ImageNet-22k classification are released.

To Do List

Introduction

RevCol is composed of multiple copies of subnetworks, named columns respectively, between which multi-level reversible connections are employed. RevCol coud serves as a foundation model backbone for various tasks in computer vision including classification, detection and segmentation.

<p align="center"> <img src="figures/title.png" width=100% height=100% class="center"> </p>

Main Results on ImageNet with Pre-trained Models

namepretrainresolution#paramsFLOPsacc@1pretrained modelfinetuned model
RevCol-TImageNet-1K224x22430M4.5G82.2baidu/github-
RevCol-SImageNet-1K224x22460M9.0G83.5baidu/github-
RevCol-BImageNet-1K224x224138M16.6G84.1baidu/github-
RevCol-B<sup>*</sup>ImageNet-22K224x224138M16.6G85.6baidu/githubbaidu/github
RevCol-B<sup>*</sup>ImageNet-22K384x384138M48.9G86.7baidu/githubbaidu/github
RevCol-L<sup>*</sup>ImageNet-22K224x224273M39G86.6baidu/githubbaidu/github
RevCol-L<sup>*</sup>ImageNet-22K384x384273M116G87.6baidu/githubbaidu/github
RevCol-H<sup>*+</sup>Megdata-168Mpretrain 224 / finetune 6402.1B253790.0huggingfacehuggingface

[+]: Note that we use a slightly different model on RevCol-H with one more branch from the bottom level to the top one. Later experiments prove that this connection is unnecessary, however, consider RevCol-H's training cost, we do not retrain it.

Getting Started

Please refer to INSTRUCTIONS.md for setting up, training and evaluation details.

Acknowledgement

This repo was inspired by several open source projects. We are grateful for these excellent projects and list them as follows:

License

RevCol is released under the Apache 2.0 license.

Contact Us

If you have any questions about this repo or the original paper, please contact Yuxuan at caiyuxuan@megvii.com.

Citation

@inproceedings{cai2022reversible,
  title={Reversible Column Networks},
  author={Cai, Yuxuan and Zhou, Yizhuang and Han, Qi and Sun, Jianjian and Kong, Xiangwen and Li, Jun and Zhang, Xiangyu},
  booktitle={International Conference on Learning Representations},
  year={2023},
  url={https://openreview.net/forum?id=Oc2vlWU0jFY}
}