Bridging The Domain Gap in Real World Super-Resolution

There has been a tremendous amount of techniques proposed to transfer artistic style from one image to another in particular techniques exploiting neural representation of data, from Convolutional Neural Networks to Generative Adversarial Networks. However, most of them do not accurately account for the semantic information related to the objects present in both images or require a considerable training set. in this paper, we want to provide a data augmentation technique as faithful as possible to the style of the target artist, while requiring as few training samples as possible (since artworks containing the same semantics from one artist are usually rare). Hence, this paper targets to improve the state-of-the-art by first apply semantic segmentation on both images to then transfer the style from the painting to a photo but preserving common semantic regions. More specifically, Van Gogh's paintings have been used for this purpose, because they are challenging to segment.