Codes for ECBSR

Edge-oriented Convolution Block for Real-time Super Resolution on Mobile Devices
Xindong Zhang, Hui Zeng, Lei Zhang
ACM Multimedia 2021

Codes

This implementation largely depends on EDSR. A lighten version will be released soon.

Dependencies & Installation

The dependencies and installation of code base can refer to EDSR. Then, move the archs and network of ECBSR to the code base of EDSR by:

mv [ECBSR]/src/option.py [EDSR]/src/
mv [ECBSR]/src/model/ecb.py [EDSR]/src/model
mv [ECBSR]/src/model/ecbsr.py [EDSR]/src/model

Training & Testing

Trained & tested on Pytorch-1.2.0. You could also try less/larger batch-size, if there are limited/enough hardware resources in your GPU-server.

cd [EDSR]/src/

## ecbsr-m4c8-x2-prelu(revise the NAME_OF_OUTPUT_FOLDER to your selected folder)
CUDA_VISIBLE_DEVICES=0 python main.py --model ECBSR --scale 2 --patch_size 128 --save NAME_OF_OUTPUT_FOLDER --reset --m_ecbsr 4 --c_ecbsr 8 --ecbsr_idt 0 --act prelu

## ecbsr-m4c8-x4-prelu
CUDA_VISIBLE_DEVICES=0 python main.py --model ECBSR --scale 4 --patch_size 256 --save NAME_OF_OUTPUT_FOLDER --reset --m_ecbsr 4 --c_ecbsr 8 --ecbsr_idt 0 --act prelu

## ecbsr-m4c16-x2-prelu
CUDA_VISIBLE_DEVICES=0 python main.py --model ECBSR --scale 2 --patch_size 128 --save NAME_OF_OUTPUT_FOLDER --reset --m_ecbsr 4 --c_ecbsr 16 --ecbsr_idt 0 --act prelu

## ecbsr-m4c16-x4-prelu
CUDA_VISIBLE_DEVICES=0 python main.py --model ECBSR --scale 4 --patch_size 256 --save NAME_OF_OUTPUT_FOLDER --reset --m_ecbsr 4 --c_ecbsr 16 --ecbsr_idt 0 --act prelu

Citation

---

@article{zhang2021edge,
  title={Edge-oriented Convolution Block for Real-time Super Resolution on Mobile Devices},
  author={Zhang, Xindong and Zeng, Hui and Zhang, Lei},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia (ACM MM)},
  year={2021}
}

Acknowledgement

This implementation largely depends on EDSR. Thanks for the excellent codebase! Our lighten version will come soon.