Voxel Deformation-Aware Neural Intersection Function

2026-04-27

AI Summary

This work addresses the challenge of efficiently and accurately predicting neural intersections for parameterized deformable and dynamic geometries without requiring retraining. By introducing a mapping mechanism between rest and deformed spaces, the method back-projects ray sample points to a canonical space, enabling a single neural network to represent geometry across diverse poses in a unified manner. The key innovations include the first integration of deformation-aware mechanisms into neural intersection functions, combined with scale-invariant distance regression, uncertainty-weighted multi-task learning, and a hybrid positional-grid encoding scheme. This approach significantly enhances rendering accuracy and generalization for dynamic scenes while preserving model compactness and inference efficiency.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

We extend the Locally-Subdivided Neural Intersection Function (LSNIF) to support parameterized deformable and animated geometry. Our approach introduces a rest-space and deformed-space formulation inspired by meshless rendering, allowing ray samples to be mapped back to a canonical space where a single neural network represents geometry consistently across poses without retraining. To maintain accuracy under deformation-aware training, we incorporate scale-invariant distance regression, uncertainty-weighted multi-task learning, and a hybrid positional-grid encoding. The resulting method preserves the compactness and efficiency of LSNIF while enabling robust neural intersection prediction for dynamic geometry.

Problem

• neural intersection
• deformable geometry
• dynamic geometry
• canonical space
• ray tracing

Innovation

• neural intersection function
• deformable geometry
• canonical space mapping
• scale-invariant distance regression
• hybrid positional-grid encoding