lin-sinorodin/swinir_wrapper

Wrapper for "SwinIR: Image Restoration Using Swin Transformer", for easy usage as a package.

Wrapper for SwinIR

Based on this repository - the official PyTorch implementation of SwinIR: Image Restoration Using Swin Transformer.

About SwinIR and this repository:

• SwinIR achieves state-of-the-art performance on six tasks: image super-resolution (including classical, lightweight and real-world image super-resolution), image denoising (including grayscale and color image denoising) and JPEG compression artifact reduction.
• This repository only provides usage of the 3 image super-resolution tasks from SwinIR at the moment.
• SwinIR.py is a minimal wrapper for the super resolution model, making it easy to use as a part from a bigger pipeline.
• Only enables the usage of the pretrained weights from the model zoo. For training, see the official repo.
• Uses the PyTorch model network_swinir.py from the official repo (unchanged).

How to Use

• For easy usage on your own data, see Demo.ipynb. This notebook will be displayed best using Google Colab which supports more interactive usage:

• As a quick preview, this example demonstrates usage with only few lines:

  import cv2
  from SwinIR_wrapper import SwinIR_SR
  
  # initialize super resolution model
  sr = SwinIR_SR(model_type='real_sr', scale=4)
  
  # load low quality image
  img_lq = cv2.imread(path, cv2.IMREAD_COLOR)
  
  # feed the image to the SR model
  img_hq = sr.upscale(img_lq)

License and Acknowledgement

Please follow the license of the official repo of this paper. Thanks for their great work!

swinir_wrapper's Issues

[REQ] HINet wrapper

Hi there, 1st of all thanks for your work, your wrapper is very clean and usable.

If you're interested can you build something similar for HINet: Half Instance Normalization Network for Image Restoration ?

With the help of HIN Block, HINet surpasses the state-of-the-art (SOTA) on various image restoration tasks. For image denoising, we exceed it 0.11dB and 0.28 dB in PSNR on SIDD dataset, with only 7.5% and 30% of its multiplier-accumulator operations (MACs), 6.8 times and 2.9 times speedup respectively. For image deblurring, we get comparable performance with 22.5% of its MACs and 3.3 times speedup on REDS and GoPro datasets. For image deraining, we exceed it by 0.3 dB in PSNR on the average result of multiple datasets with 1.4 times speedup.

Some "real-world" tests by Selur:

Mode: Deblur GoPro

Mode: Deblur REDS

Mode: denoise

Mode: derain

Hope that inspires !