SRCNN-Reimplementation-on-DIV2K-Datasets

This repository contains a reimplementation of the SRCNN (Super-Resolution Convolutional Neural Network) for image super-resolution, trained and evaluated on DIV2K Datasets.
It is a PyTorch-based implementation that demonstrates high-quality upscaling and performance evaluation with PSNR and SSIM metrics.

Models

You can download trained SRCNN models (trained on DIV2K (x3 only)) here:

• SRCNN x3 (Google Drive)

Features

• Reimplementation of the original SRCNN architecture
• Supports training and testing on DIV2K Datasets
• Evaluation with PSNR and SSIM metrics
• Save and load model checkpoints
• Example scripts for training and inference

📊 Example Outputs

Left: Baseline SRCNN    |    Right: Improved SRCNN