«RotNet» realizes image rotation correction based on deep learning

• Table of Contents
• Background
• Installation
• Usage
  • How to use
  • How to add dataset
• Maintainers
• Thanks
• Contributing
• License

Looking for information on the Internet, we found that the image rotation angle can be detected by deep learning algorithm. Refer to

• d4nst/RotNet
• Correcting Image Orientation Using Convolutional Neural Networks
• Image Orientation Estimation with Convolutional Networks
• UNSUPERVISED REPRESENTATION LEARNING BY PREDICTING IMAGE ROTATIONS

The corresponding implementation can't meet the current performance requirements, so I implement one myself

$ pip install -r requirements.txt

• train

$ export PYTHONPATH=<root path>
$ CUDA_VISIBLE_DEVICES=0 python tools/train.py -cfg=configs/xxx.yaml

• test

$ export PYTHONPATH=<root path>
$ CUDA_VISIBLE_DEVICES=0 python demo/demo.py -cfg=demo/xxx.yaml

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)

• zhujian - Initial work - zjykzj

• d4nst/RotNet
• ZJCV/ZCls

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.

Apache License 2.0 © 2020 zjykzj