MARF: The Medial Atom Ray Field Object Representation
arXiv:2307.00037 · doi:10.1016/j.cag.2023.06.032
Abstract
We propose Medial Atom Ray Fields (MARFs), a novel neural object representation that enables accurate differentiable surface rendering with a single network evaluation per camera ray. Existing neural ray fields struggle with multi-view consistency and representing surface discontinuities. MARFs address both using a medial shape representation, a dual representation of solid geometry that yields cheap geometrically grounded surface normals, in turn enabling computing analytical curvature despite the network having no second derivative. MARFs map a camera ray to multiple medial intersection candidates, subject to ray-sphere intersection testing. We illustrate how the learned medial shape quantities applies to sub-surface scattering, part segmentation, and aid representing a space of articulated shapes. Able to learn a space of shape priors, MARFs may prove useful for tasks like shape retrieval and shape completion, among others. Code and data can be found at https://github.com/pbsds/MARF.
arXiv commentTo be published in 3DOR 2023 and C&G Volume 114
References in corpus (21)
- Adam: A Method for Stochastic Optimization
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Instant Neural Graphics Primitives with a Multiresolution Hash Encoding
- The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
- Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains
- SurfaceNet: Adversarial SVBRDF Estimation from a Single Image
- DONeRF: Towards Real-Time Rendering of Compact Neural Radiance Fields using Depth Oracle Networks
- DISN: Deep Implicit Surface Network for High-quality Single-view 3D Reconstruction
- MetaSDF: Meta-learning Signed Distance Functions
- ReLU Fields: The Little Non-linearity That Could
- HyperNeRF: A Higher-Dimensional Representation for Topologically Varying Neural Radiance Fields
- Object-Centric Neural Scene Rendering
- Animatable Neural Radiance Fields from Monocular RGB Videos
- SEG-MAT: 3D Shape Segmentation Using Medial Axis Transform
- Adversarial Generation of Continuous Implicit Shape Representations
- Exploring Differential Geometry in Neural Implicits
- LSMAT Least Squares Medial Axis Transform
- Is Attention All That NeRF Needs?
- Differentiable Rendering with Reparameterized Volume Sampling
- nerf2nerf: Pairwise Registration of Neural Radiance Fields
- FIRe: Fast Inverse Rendering using Directional and Signed Distance Functions