SRGAN: Training Dataset Matters
This work addresses dataset dependency in super-resolution for computer vision applications, but it is incremental as it builds on existing SRGAN architecture.
The study investigates how training dataset selection affects SRGAN performance, finding that networks learn to reconstruct objects rather than just sharpen edges, and achieve better Frechet Inception Distance scores when trained on the same dataset as inference images.
Generative Adversarial Networks (GANs) in supervised settings can generate photo-realistic corresponding output from low-definition input (SRGAN). Using the architecture presented in the SRGAN original paper [2], we explore how selecting a dataset affects the outcome by using three different datasets to see that SRGAN fundamentally learns objects, with their shape, color, and texture, and redraws them in the output rather than merely attempting to sharpen edges. This is further underscored with our demonstration that once the network learns the images of the dataset, it can generate a photo-like image with even a slight hint of what it might look like for the original from a very blurry edged sketch. Given a set of inference images, the network trained with the same dataset results in a better outcome over the one trained with arbitrary set of images, and we report its significance numerically with Frechet Inception Distance score [22].