LSNIF: Locally-Subdivided Neural Intersection Function

2025-04-30

AI Summary

This work addresses the high memory overhead and poor generalizability of traditional BVHs in ray tracing. We propose a neural implicit intersection modeling approach that replaces explicit geometric acceleration structures with locally refined neural intersection functions, enabled by sparse hash grid encoding and geometric voxelization for efficient spatial representation. Our method employs a scene-agnostic, single-object offline training paradigm and introduces a multi-task loss function jointly optimizing hit-point localization, visibility classification, and material index prediction. Crucially, it preserves full view independence and full-path tracing capability. Evaluated on real-world rendering scenes, our approach achieves up to 106.2× memory compression over conventional BVH-based ray tracers, significantly enhancing storage efficiency and deployment flexibility of neural rendering pipelines.

Technology Category

• Computer Vision: 3D Computer Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• 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
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Neural representations have shown the potential to accelerate ray casting in a conventional ray-tracing-based rendering pipeline. We introduce a novel approach called Locally-Subdivided Neural Intersection Function (LSNIF) that replaces bottom-level BVHs used as traditional geometric representations with a neural network. Our method introduces a sparse hash grid encoding scheme incorporating geometry voxelization, a scene-agnostic training data collection, and a tailored loss function. It enables the network to output not only visibility but also hit-point information and material indices. LSNIF can be trained offline for a single object, allowing us to use LSNIF as a replacement for its corresponding BVH. With these designs, the network can handle hit-point queries from any arbitrary viewpoint, supporting all types of rays in the rendering pipeline. We demonstrate that LSNIF can render a variety of scenes, including real-world scenes designed for other path tracers, while achieving a memory footprint reduction of up to 106.2x compared to a compressed BVH.

Problem

• Replaces BVHs with neural networks for ray casting acceleration
• Enables neural output of visibility, hit-points, and material indices
• Reduces memory usage by 106.2x versus compressed BVH

Innovation

• Replaces BVHs with neural network for rendering
• Uses sparse hash grid encoding scheme
• Outputs visibility, hit-points, and material indices