zcls 0.4.0

Object Classification Code Base

Language: 🇺🇸 🇨🇳

«ZCls» is a classification model benchmark code base

Supported Recognizers:

Refer to roadmap for details

Table of Contents

• Table of Contents
• Background
• Usage
  • Installation
  • How to Use
  • How to add Dataset
• Maintainers
• Thanks
• Contributing
• License

Background

In order to further improve the algorithm performance, it is usually necessary to improve the existing model, which inevitably involves code refactoring. Creating this repo, on the one hand, serves as the CodeBase of the new model/optimization method, on the other hand, it also records the comparison between the custom model and the existing implementation (such as Torchvision Models)

Usage

Installation

$ pip install zcls

How to Use

1. Add dataset path to config_file, like CIFAR100

  NAME: 'CIFAR100'
  TRAIN_ROOT: './data/cifar'
  TEST_ROOT: './data/cifar'

Note: current support CIFAR10/CIFAR100/FashionMNIST/ImageNet

2. Add environment variable

$ export PYTHONPATH=/path/to/ZCls

3. Train

$ CUDA_VISIBLE_DEVICES=0 python tool/train.py -cfg=configs/benchmarks/r50_cifar100_224_e100_rmsprop.yaml

After training, the corresponding model can be found in outputs/, add model path to xxx.yaml

    PRELOADED: ""

4. Test

$ CUDA_VISIBLE_DEVICES=0 python tool/test.py -cfg=configs/benchmarks/r50_cifar100_224_e100_rmsprop.yaml

5. If finished the training halfway, resume it like this

$ CUDA_VISIBLE_DEVICES=0 python tool/train.py -cfg=configs/benchmarks/r50_cifar100_224_e100_rmsprop.yaml --resume

6. Use multiple GPU to train

$ CUDA_VISIBLE_DEVICES=0<,1,2,3> python tool/train.py -cfg=configs/benchmarks/r50_cifar100_224_e100_rmsprop.yaml -g=<N>

How to add Dataset

Suppose your dataset is in the following format

root/dog/xxx.png
root/dog/xxy.png
root/dog/xxz.png

root/cat/123.png
root/cat/nsdf3.png
root/cat/asd932_.png

modify config_file like this

DATASET:
  NAME: 'GeneralDataset'
  TRAIN_ROOT: /path/to/train_root
  TEST_ROOT: /path/to/test_root
  TOP_K: (1, 5)

Maintainers

• zhujian - Initial work - zjykzj

Thanks

@misc{ding2021repvgg,
      title={RepVGG: Making VGG-style ConvNets Great Again}, 
      author={Xiaohan Ding and Xiangyu Zhang and Ningning Ma and Jungong Han and Guiguang Ding and Jian Sun},
      year={2021},
      eprint={2101.03697},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

@misc{fan2020pyslowfast,
  author =       {Haoqi Fan and Yanghao Li and Bo Xiong and Wan-Yen Lo and
                  Christoph Feichtenhofer},
  title =        {PySlowFast},
  howpublished = {\url{https://github.com/facebookresearch/slowfast}},
  year =         {2020}
}

@misc{zhang2020resnest,
      title={ResNeSt: Split-Attention Networks}, 
      author={Hang Zhang and Chongruo Wu and Zhongyue Zhang and Yi Zhu and Haibin Lin and Zhi Zhang and Yue Sun and Tong He and Jonas Mueller and R. Manmatha and Mu Li and Alexander Smola},
      year={2020},
      eprint={2004.08955},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

@misc{ding2019acnet,
      title={ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks}, 
      author={Xiaohan Ding and Yuchen Guo and Guiguang Ding and Jungong Han},
      year={2019},
      eprint={1908.03930},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

@misc{howard2019searching,
      title={Searching for MobileNetV3}, 
      author={Andrew Howard and Mark Sandler and Grace Chu and Liang-Chieh Chen and Bo Chen and Mingxing Tan and Weijun Wang and Yukun Zhu and Ruoming Pang and Vijay Vasudevan and Quoc V. Le and Hartwig Adam},
      year={2019},
      eprint={1905.02244},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

@misc{cao2019gcnet,
      title={GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond}, 
      author={Yue Cao and Jiarui Xu and Stephen Lin and Fangyun Wei and Han Hu},
      year={2019},
      eprint={1904.11492},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

For more thanks, check THANKS

Contributing

Anyone's participation is welcome! Open an issue or submit PRs.

Small note:

• Git submission specifications should be complied with Conventional Commits
• If versioned, please conform to the Semantic Versioning 2.0.0 specification
• If editing the README, please conform to the standard-readme specification.

License

Apache License 2.0 © 2020 zjykzj