This repository contains the implementation and inference code for EfRLFN, a deep learning model for single image super-resolution.

• Efficient residual architecture for high-quality image upscaling
• Supports multiple upscaling factors
• Fast inference on CUDA-enabled GPUs

pip install -r requirements.txt

To upscale an image using a trained model:

python inference.py -w [WEIGHTS_PATH] -s [SCALE_FACTOR] -i [INPUT_IMAGE] -o [OUTPUT_IMAGE]

-w/--weights: Path to the pretrained model weights (.pt or .ckpt format)

-s/--scale: Upscaling factor (e.g., 2, 4)

-i/--input: Path to input image

-o/--output: Path to save the output image

python inference.py -w weights/EfRLFN-4x-model.ph -s 4 -i images/low_res.jpg -o images/high_res.jpg

Pretrained weights are available for different scale factors:

EfRLFN x2

EfRLFN x4

The proposed dataset can be downloaded here:

Train dataset

Test dataset