Keras implementation of "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"

The generator creates a high-resolution (HR) image (4x upscaled) from a corresponding low-resolution (LR) image. The discriminator distinguishes the generated (fake) HR images from the original HR images.

Figure 4 from paper: Architecture of Generator and Discriminator Network with corresponding kernel size (k), number of feature maps (n) and stride (s) indicated for each convolutional layer.

Code Overview: Overview of the three networks; generator, discriminator, and VGG19. Generator create SR image from LR, discriminator predicts whether it's a SR or original HR, and VGG19 extracts features from generated SR and original HR images.

Losses Overview: The perceptual loss is a combination of content loss (based on VGG19 features) and adversarial loss. Equations are taken directly from "original paper".

A command-line interface can be found in train.py. To train run e.g.:

python train.py \
    --train <TRAINING_IMAGES_PATH> \
    --validation <VALIDATION_IMAGES_PATH> \
    --scale 4 \
    --test_path images/samples_4X \
    --stage all

Check the example_usage notebook: example_usage.ipynb