Automatic Differentiable Procedural Modeling

• DOI: 10.1111/cgf.14475
• Publication Date: 2022-05-25T06:25:20Z
• Authors: Mathieu Gaillard; Vojtěch Krs; Giorgio Gori; Radomír Měch; Bedrich Benes
• Keywords: Procedural modeling; Representation

Abstract Procedural modeling allows for an automatic generation of large amounts similar assets, but there is limited control over the generated output. We address this problem by introducing Automatic Differentiable Modeling (ADPM). The forward procedural model generates a final editable model. user modifies output interactively, and modifications are transferred back to as its parameters solving inverse problem. present auto‐differentiable representation that significantly accelerates optimization. In ADPM always available, all changes non‐destructive, can interactively 3D object while keeping representation. provides with precise resulting comparable non‐procedural interactive modeling. node‐based, it hierarchical scene geometry converted differentiable computational graph. Our formulation focuses on differentiability high‐level primitives bounding volumes components rather than detailed mesh geometry. Although limits expressiveness edits, efficient derivative computation enables interactivity. designed new optimizer solve It detect edit under‐determined has degrees freedom. Leveraging cheap evaluation, explore region optimality edits suggest various configurations, which achieve requested differently. show our system's efficiency several examples, we validate study.