Signed Distance Fields (SDFs) parameterized by neural networks have recently gained popularity as a fundamental geometric representation. However, editing the shape encoded by a neural SDF remains an open challenge. A tempting approach is to leverage common geometric operators (e.g., boolean operations), but such edits often lead to incorrect non-SDF outputs (which we call Pseudo-SDFs), preventing them from being used for downstream tasks. In this paper, we characterize the space of Pseudo-SDFs, which are eikonal yet not true distance functions, and derive the closest point loss, a novel regularizer that encourages the output to be an exact SDF. We demonstrate the applicability of our regularization to many operations in which traditional methods cause a Pseudo-SDF to arise, such as CSG and swept volumes, and produce a true (neural) SDF for the result of these operations.
CITATION STYLE
Marschner, Z., Sellán, S., Liu, H. T. D., & Jacobson, A. (2023). Constructive Solid Geometry on Neural Signed Distance Fields. In Proceedings - SIGGRAPH Asia 2023 Conference Papers, SA 2023. Association for Computing Machinery, Inc. https://doi.org/10.1145/3610548.3618170
Mendeley helps you to discover research relevant for your work.