MLP-SRGAN: A Single-Dimension Super Resolution GAN using MLP-Mixer

Dependencies

• Python 3
• Python packages: pip install numpy scikit-image scipy PyWavelets pandas torch torchvision einops matplotlib nibabel basicsr

Command Line Usage

Ideal input image size is 256 x 64, tiling will be used if images exceed these dimensions.

Usage: python inference_mlpsrgan.py -n mlp-srgan-d-1 -i infile -o outfile [options]...

  -h                   show this help
  -i --input           Input image or folder | for 3D medical images use axial plane. Default: inputs
  -o --output          Output folder. Default: results
  -n --model_name      Model name. Default: RealESRGAN_x4plus
  -s, --outscale       The final upsampling scale of the image (only 4 is available at the moment). Default: 4
  --suffix             Suffix of the restored image. Default: out
  -t, --tile           Tile size, 0 for no tile during testing. Default: 0
  --fp16               Use fp16 precision during inference. Default: fp32 (single precision).
  --ext                Image extension. Options: auto | jpg | png, auto means using the same extension as inputs. Default: auto

Pretrained Models

Pretrained models are available on Google Drive at the following link: https://drive.google.com/drive/folders/1q4f1Yzraqtgdplw9dAtbdtWLGSm7vzHx?usp=sharing

Model Diagrams

Image Samples

Contact

If you have any questions please email [email protected].

License

Citations

See Also

This repository uses the PyTorch MLP-Mixer.

This repository uses the format provided by BasicSR. Please check out the repository!

This work is inspired by Real-ESRGAN for natural images.