Shape Synthesis

Shape synthesis focuses on computationally generating 3D shapes, aiming to create realistic and diverse models from various inputs, including user sketches or point clouds. Current research emphasizes part-based representations, leveraging neural implicit functions and generative adversarial networks (GANs) to control shape details and topology, often incorporating techniques like medial axis transforms and differentiable rendering. These advancements are improving the fidelity and controllability of synthesized shapes, with applications in areas such as computer-aided design, 3D printing, and virtual/augmented reality.