diwu121/openmixup

CAIRI Supervised, Semi- and Self-Supervised Visual Learning Toolbox and Benchmark

OpenMixup

📘Documentation | 🛠️Installation | 🚀Model Zoo | 👀Awesome Mixup | 🆕News

Introduction

The main branch works with PyTorch 1.8 (required by some self-supervised methods) or higher (we recommend PyTorch 1.10). You can still use PyTorch 1.6 for supervised classification methods.

OpenMixup is an open-source toolbox for supervised, self-, and semi-supervised visual representation learning with mixup based on PyTorch, especially for mixup-related methods.

Major Features

• Modular Design. OpenMixup follows a similar code architecture of OpenMMLab projects, which decompose the framework into various components, and users can easily build a customized model by combining different modules. OpenMixup is also transplatable to OpenMMLab projects (e.g., MMSelfSup).

• All in One. OpenMixup provides popular backbones, mixup methods, and self-supervised algorithms. Users can perform supervised training or self-supervised pre-training under the same setting.

• Standard Benchmarks. OpenMixup supports standard benchmarks of image classification, mixup classification, self-supervised evaluation, and provides smooth evaluation on downstream tasks with open-source projects (e.g., object detection and segmentation on Detectron2 and MMSegmentation).

What's New

[2020-07-07] OpenMixup v0.2.4 is released (issue #7), which fixs bugs of weight initialization and fine-tuning, updates docs, etc.

Installation

There are quick installation steps for develepment:

conda create -n openmixup python=3.8 pytorch=1.10 cudatoolkit=11.3 torchvision -c pytorch -y
conda activate openmixup
pip3 install openmim
mim install mmcv-full
git clone https://github.com/Westlake-AI/openmixup.git
cd openmixup
python setup.py develop

Please refer to install.md for more detailed installation and dataset preparation.

Getting Started

Please see get_started.md for the basic usage of OpenMixup. You can start a multiple GPUs training with CONFIG_FILE using the following script. An example,

bash tools/dist_train.sh ${CONFIG_FILE} ${GPUS} [optional arguments]

Please Then, see Tutorials for more tech details:

• config files
• add new dataset
• data pipeline
• add new modules
• customize schedules
• customize runtime
• benchmarks

Overview of Model Zoo

Please refer to Model Zoos for various backbones, mixup methods, and self-supervised algorithms. We also provide the paper lists of Awesome Mixups for your reference. Checkpoints and traning logs will be updated soon!

• Backbone architectures for supervised image classification on ImageNet.

  Currently supported backbones

  • VGG (ICLR'2015) [readme]
  • ResNet (CVPR'2016) [readme]
  • ResNeXt (CVPR'2017) [readme]
  • SE-ResNet (CVPR'2018) [readme]
  • SE-ResNeXt (CVPR'2018) [readme]
  • ShuffleNetV2 (ECCV'2018) [readme]
  • MobileNetV2 (CVPR'2018) [readme]
  • MobileNetV3 (ICCV'2019)
  • EfficientNet (ICML'2019) [readme]
  • Swin-Transformer (ICCV'2021) [readme]
  • RepVGG (CVPR'2021)
  • Vision-Transformer (ICLR'2021) [readme]
  • MLP-Mixer (NIPS'2021) [readme]
  • DeiT (ICML'2021)
  • ConvMixer (Openreview'2021) [readme]
  • UniFormer (ICLR'2022) [readme]
  • PoolFormer (CVPR'2022) [readme]
  • ConvNeXt (CVPR'2022) [readme]
  • VAN (ArXiv'2022) [readme]

• Mixup methods for supervised image classification.

  Currently supported mixup methods

  • Mixup (ICLR'2018)
  • CutMix (ICCV'2019)
  • ManifoldMix (ICML'2019)
  • FMix (ArXiv'2020)
  • AttentiveMix (ICASSP'2020)
  • SmoothMix (CVPRW'2020)
  • SaliencyMix (ICLR'2021)
  • PuzzleMix (ICML'2020)
  • GridMix (Pattern Recognition'2021)
  • ResizeMix (ArXiv'2020)
  • AutoMix (ECCV'2022) [readme]
  • SAMix (ArXiv'2021) [readme]
  Currently supported datasets for mixups

  • ImageNet [download] [readme]
  • CIFAR-10 [download] [readme]
  • CIFAR-100 [download] [readme]
  • Tiny-ImageNet [download] [readme]
  • CUB-200-2011 [download] [readme]
  • FGVC-Aircraft [download] [readme]
  • StandfoldCars [download]
  • Place205 [download] [readme]
  • iNaturalist-2017/2018 [download] [readme]

• Self-supervised algorithms for visual representation.

  Currently supported self-supervised algorithms

  • Relative Location (ICCV'2015) [readme]
  • Rotation Prediction (ICLR'2018) [readme]
  • DeepCluster (ECCV'2018) [readme]
  • NPID (CVPR'2018) [readme]
  • ODC (CVPR'2020) [readme]
  • MoCov1 (CVPR'2020) [readme]
  • SimCLR (ICML'2020) [readme]
  • MoCoV2 (ArXiv'2020) [readme]
  • BYOL (NIPS'2020) [readme]
  • SwAV (NIPS'2020) [readme]
  • DenseCL (CVPR'2021) [readme]
  • SimSiam (CVPR'2021) [readme]
  • Barlow Twins (ICML'2021) [readme]
  • MoCoV3 (ICCV'2021) [readme]
  • MAE (CVPR'2022) [readme]
  • SimMIM (CVPR'2022) [readme]
  • CAE (ArXiv'2022)
  • A2MIM (ArXiv'2022) [readme]

Change Log

Please refer to changelog.md for details and release history.

License

This project is released under the Apache 2.0 license.

Acknowledgement

• OpenMixup is an open-source project for mixup methods created by researchers in CAIRI AI LAB. We encourage researchers interested in visual representation learning and mixup methods to contribute to OpenMixup!
• This repo borrows the architecture design and part of the code from MMSelfSup and MMClassification.

Citation

If you find this project useful in your research, please consider cite our repo:

@misc{2022openmixup,
    title={{OpenMixup}: Open Mixup Toolbox and Benchmark for Visual Representation Learning},
    author={Li, Siyuan and Liu, Zichen and Wu, Di and Stan Z. Li},
    howpublished = {\url{https://github.com/Westlake-AI/openmixup}},
    year={2022}
}

Contributors

For now, the direct contributors include: Siyuan Li (@Lupin1998), Zicheng Liu (@pone7), and Di Wu (@wudi-bu). We thanks contributors for MMSelfSup and MMClassification.

Contact

This repo is currently maintained by Siyuan Li ([email protected]) and Zicheng Liu ([email protected]).