Creative Sketch Generation
This work provides a new benchmark and method for generating creative sketches, addressing a gap in existing sketch generation research which primarily focuses on mundane sketches, benefiting artists and creative tool developers.
This paper addresses the generation of creative sketches by introducing two new datasets, Creative Birds and Creative Creatures, each with 10k sketches and part annotations. They propose DoodlerGAN, a part-based Generative Adversarial Network, which generates sketches that are quantitatively and qualitatively more creative and higher quality than existing methods, with human subjects preferring DoodlerGAN's output over human-drawn sketches for Creative Birds.
Sketching or doodling is a popular creative activity that people engage in. However, most existing work in automatic sketch understanding or generation has focused on sketches that are quite mundane. In this work, we introduce two datasets of creative sketches -- Creative Birds and Creative Creatures -- containing 10k sketches each along with part annotations. We propose DoodlerGAN -- a part-based Generative Adversarial Network (GAN) -- to generate unseen compositions of novel part appearances. Quantitative evaluations as well as human studies demonstrate that sketches generated by our approach are more creative and of higher quality than existing approaches. In fact, in Creative Birds, subjects prefer sketches generated by DoodlerGAN over those drawn by humans! Our code can be found at https://github.com/facebookresearch/DoodlerGAN and a demo can be found at http://doodlergan.cloudcv.org.