4SKST

This is 4 sketch style (4SKST) dataset, from the research paper "Semi-supervised reference-based sketch extraction using a contrastive learning framework" Dataset consists one of four different styles of sketches paired to color images.

Total 25 color images and corresponding 100 (25x4) sketches.

Sketches are drawn by professional artist.

Base color images are mainly from https://jewel-s.jp/ (Game source for free share), Flickr (CC-2.0), and generated by deep-learning model (StyleGAN, stablediffusion)