Vertex Features for Neural Global Illumination

2025-08-11

AI Summary

Traditional voxel-based feature grids suffer from excessive memory consumption, hindering neural rendering efficiency. To address this, we propose Neural Vertex Features—a novel representation that relocates learnable features from volumetric grids to vertices of an explicit triangle mesh for the first time. Leveraging surface geometry as a structural prior, our method aligns feature distributions with object surfaces, yielding a compact and semantically consistent representation. Our framework jointly optimizes explicit mesh geometry, neural radiance fields, and task-driven geometric constraints—enabling high-fidelity rendering under complex lighting conditions such as global illumination. Experiments demonstrate that our approach reduces memory usage to less than 20% of conventional grid-based methods, significantly lowers inference overhead, and maintains competitive rendering quality. The core contribution lies in redefining feature storage around geometric primitives—using mesh vertices as anchors—to break the memory-accuracy trade-off inherent in prior implicit or volumetric representations.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Representation Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

Recent research on learnable neural representations has been widely adopted in the field of 3D scene reconstruction and neural rendering applications. However, traditional feature grid representations often suffer from substantial memory footprint, posing a significant bottleneck for modern parallel computing hardware. In this paper, we present neural vertex features, a generalized formulation of learnable representation for neural rendering tasks involving explicit mesh surfaces. Instead of uniformly distributing neural features throughout 3D space, our method stores learnable features directly at mesh vertices, leveraging the underlying geometry as a compact and structured representation for neural processing. This not only optimizes memory efficiency, but also improves feature representation by aligning compactly with the surface using task-specific geometric priors. We validate our neural representation across diverse neural rendering tasks, with a specific emphasis on neural radiosity. Experimental results demonstrate that our method reduces memory consumption to only one-fifth (or even less) of grid-based representations, while maintaining comparable rendering quality and lowering inference overhead.

Problem

• Reducing memory footprint in neural rendering
• Improving feature representation with mesh vertices
• Maintaining rendering quality with lower memory

Innovation

• Stores learnable features at mesh vertices
• Leverages geometry for compact representation
• Reduces memory use while maintaining quality