Creative Sketch Generation
Songwei Ge, Vedanuj Goswami, C. Lawrence Zitnick, Devi Parikh

TL;DR
This paper introduces new datasets and a novel GAN-based model, DoodlerGAN, for generating highly creative and high-quality sketches, outperforming existing methods and even surpassing human-drawn sketches in some cases.
Contribution
The work presents two new datasets of creative sketches and a part-based GAN model that produces more creative and higher-quality sketches than previous approaches.
Findings
Generated sketches are preferred over human sketches in some cases.
DoodlerGAN outperforms existing sketch generation methods.
Human studies confirm the high quality and creativity of generated sketches.
Abstract
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…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
Taxonomy
TopicsComputer Graphics and Visualization Techniques · 3D Shape Modeling and Analysis · Human Motion and Animation
