GRASS Image (IMAGE) Association for Computing Machinery Caption GRASS, a Generative Recursive Autoencoder for Shape Structures, enables generating novel and structurally plausible 3-D shapes. Shown here, an example for generating in-betweens given two chairs: discrete blending of translational symmetries (slats on the chair backs) and rotational symmetries (the swivel legs). GRASS separately encodes and synthesizes box structures (bottom) and part geometries (top). The blending is data-driven and learned by the unsupervised autoencoder. Credit National University of Defense Technology Usage Restrictions None License Licensed content Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.