Image Upscaling with a Super-Resolution Convolutional Neural Network

This project implements a Super-Resolution Convolutional Neural Network (SRCNN) for image upscaling.

Examples

Low Resolution 0X	High Resolution 6x

Requirements

• Python 3.x
• PyTorch
• torchvision
• tqdm
• matplotlib

Usage

1. Prepare your data in newdata/low and newdata/high directories.
2. Run training: python train_newdata.py
3. For inference: python inference.py

Files

• model.py: SRCNN model definition
• train_newdata.py: Training script for new dataset
• inference.py: Inference script
• data_loader_multiscale.py: Data loader for multi-scale images

Dataset

• https://www.kaggle.com/code/quadeer15sh/image-super-resolution-using-autoencoders/input