MARF: The Medial Atom Ray Field Object Representation

TL;DR

MARFs introduce a neural object representation that accurately renders surfaces with a single ray evaluation, effectively handling surface discontinuities and enabling applications like shape segmentation and articulation.

Contribution

The paper presents Medial Atom Ray Fields (MARFs), a novel medial shape-based neural representation that improves multi-view consistency and surface normal computation in neural rendering.

Findings

Enables accurate differentiable surface rendering with a single network evaluation per ray.

Addresses multi-view consistency and surface discontinuities effectively.

Supports applications like sub-surface scattering, part segmentation, and articulated shape modeling.

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.