pseudo-sr-main

CycleGAN-Transformer-SR README:

• To check quota and file limit for this shared directory:

  /scratch/group/mburrisgroup/check_quota.mburrisgroup.sh
  

• To fix or reset permission in this shared directory (make all files and dirs readable/writable by the group; must run by the owner of files/dirs whose permssion needs to be fixed):

  /scratch/group/mburrisgroup/fix_permission.mburrisgroup.sh
  

Introduction

This repo is based on pseudo-sr.

You can get the dataset from here. After unzip, put HR dataset and LR dataset in the Dataset folder "/../Dataset/LOW/LR" and "/../Dataset/HIGH/HR", respectively. Make sure the data path configured in Train.py file is the same where it is located.

The integrated Transformer model is based on Restomrer (https://github.com/swz30/Restormer)

Usage

First, configure the yaml file which is located at configs/faces.yaml. Set the root folder of face dataset to DATA.FOLDER.

After you download the dataset, please save the high-resolution images to /../Dataset/HIGH/HR, and save the low-resolution images to ../../Dataset/LOW/LR. Make sure the dataset path is the same to the directory configure in Train.py

Environment: Python (Pytorch)

To train: python3 train.py configs/faces.yaml

CUDA_VISIBLE_DEVICES=2,3 python3 train.py configs/faces.yaml --port 12121

The --port option is only required for multi-gpu training. You can use a number between 49152 and 65535 for the port number.

Reference

Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang. Restormer: Efficient Transformer for High-Resolution Image Restoration

Restormer-offocial-codes

pseudo-sr-instance-codes.