SRGAN-from-scratch

My implementation of the model from the paper "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"

I trained the model on the DIV2K-dataset

Instal dependencies:

pip install -r requirements.txt

To train using your custom dataset:
Clone the repository:

git clone https://github.com/pbcong/SRGAN-from-scratch

Train using train-notebook or run train.py

Pretrained models:

• With VGG19 and MSE only: Here
• SRGAN: Here

Results:\