Abstract
The study of neural generative models of human sketches is a fascinatingcontemporary modeling problem due to the links between sketch image generationand the human drawing process. The landmark SketchRNN provided breakthrough bysequentially generating sketches as a sequence of waypoints. However this leadsto low-resolution image generation, and failure to model long sketches. In thispaper we present B\'ezierSketch, a novel generative model for fully vectorsketches that are automatically scalable and high-resolution. To this end, wefirst introduce a novel inverse graphics approach to stroke embedding thattrains an encoder to embed each stroke to its best fit B\'ezier curve. Thisenables us to treat sketches as short sequences of paramaterized strokes andthus train a recurrent sketch generator with greater capacity for longersketches, while producing scalable high-resolution results. We reportqualitative and quantitative results on the Quick, Draw! benchmark.