In plain words: A neural network learns to draw crude sketches of common objects one stroke at a time, trained on thousands of human doodles across hundreds of categories. Unlike pixel-by-pixel image generators, it produces coherent drawings made of lines defined by coordinates, not pixels.
Abstract
We present sketch-rnn, a recurrent neural network (RNN) able to construct stroke-based drawings of common objects. The model is trained on thousands of crude human-drawn images representing hundreds of classes. We outline a framework for conditional and unconditional sketch generation, and describe new robust training methods for generating coherent sketch drawings in a vector format.
David Ha, Douglas Eck
arXiv:1704.03477 · cs.NE, cs.LG, stat.ML · submitted Apr 11, 2017 · updated May 19, 2017
abstract · pdf · html
https://twitter.com/hardmaru/status/852312400481079296