This repository is an implementation of EDSR model using PyTorch Check the official code and paper (The video input/output super-resolution will be added soon to the Master branch)
This model was trained on NVIDIA P100 with ~4500 annotated satellite Low Resolution image pathches which were obtained by Image degradation and Image downsampling of the High Resolution images. You can download the dataset which I used for training the model here
- Python 3.6
- PyTorch >= 1.0.0
- Pillow
- Utility
- Imageio
- tqdm
- scikit-image
- OpenCV (Only for Video input/output)
- matplotlib
You can test the pretrained model EDSR_baseline_x2 by placing the images in test folder. The supported formats are png and jpeg files. The pretrained model can be downloaded using this link
Run the script in src folder. Uncomment the appropriate line in demo.sh to test your image.
sh demo.sh
The output images can be found under experiment/test/results-Demo folder.
