J-RESRGAN

This is a fork of https://github.com/xinntao/Real-ESRGAN with some modifications to perform super-resolution enhancement specifically for J-Resolved NMR spectral data.

Dependencies, Installation & Usage

• Python >= 3.7 (Recommend to use Anaconda or Miniconda)
• PyTorch >= 1.7

Installation

1. Clone repo

   git clone https://github.com/yanyan5420/J-RESRGAN.git
   cd J-RESRGAN

2. Install dependent packages

   # Install the modified basicsr - https://github.com/yanyan5420/BasicSR
   # We use BasicSR for both training and inference
   pip install git+https://github.com/yanyan5420/BasicSR.git
   pip install -r requirements.txt
   python setup.py develop

Inference J-Res Spectra

1. Download pre-trained model: JRESRGAN_x2plus.pth

wget https://github.com/yanyan5420/J-RESRGAN/releases/download/v1.0.0/JRESRGAN_x2plus.pth -P weights

2. Inference: First, put a J-Res spectrum into inputs folder; and run the command:

python inference_realesrgan.py -n JRESRGAN_x2plus -i inputs -s 2

Common command options:
  -h                   show this help
  -i --input           Input image or folder. Default: inputs
  -o --output          Output folder. Default: results
  -n --model_name      Model name. Default: JRESRGAN_x2plus
  -s, --outscale       The final upsampling scale of the image. Default: 2
  -t, --tile           Tile size, 0 for no tile during testing. Default: 0
  --fp32               Use fp32 precision during inference. Default: fp16 (half precision).

Outputs will show in the results folder.

Online Inference

You can use the Online Inference Demo to get a quick inference.

Contact

If you have any question, please email [email protected].

🤗 Acknowledgement