Neural Network BVH Rebuild for Dynamic Mesh Ray Tracing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Building a bounding volume hierarchy (BVH) for dynamic geometries in ray tracing is an expensive process, especially in applications where scene geometry changes frequently, such as in video games or real-time simulations.
Innovation Solution
Utilizing trained neural networks to generate BVHs from meshes, with each network corresponding to a specific mesh topology, allowing generation of BVHs for various animation states of a scene without the need for complex calculations at runtime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional BVH construction algorithms are used for dynamic geometries, then accurate bounding volume hierarchy is achieved, but computational cost and reconstruction time increase significantly
Solution Approach 1:
The system performs preliminary action by training neural networks offline on representative mesh datasets before runtime. The trained networks are then reused during animation to generate BVHs instantly without complex calculations, resolving the contradiction between accuracy and speed by shifting computational burden to the training phase.
Solution Approach 2:
The patent creates simplified copies of the mesh topology with associated neural network models that can rapidly generate BVH structures. These network-based copies approximate the complex geometric relationships without requiring full traditional BVH construction algorithms, maintaining sufficient accuracy while dramatically reducing computation time.
2Reliability
If traditional BVH reconstruction is performed for each animation state, then up-to-date bounding volumes are maintained, but computational overhead becomes prohibitive for real-time applications
Solution Approach 1:
The patent replaces the mechanical/computational system of traditional BVH construction algorithms with a neural network-based system. The neural networks, trained offline, substitute for the expensive runtime calculations, enabling real-time BVH generation for animated meshes with minimal computational overhead while maintaining currency with animation states.
3Reliability
If complex BVH construction algorithms are used, then robust BVH generation is achieved, but runtime performance deteriorates
Solution Approach 1:
The system performs the complex and robust BVH construction work in advance during the offline training phase. Multiple sample meshes are processed through traditional algorithms to create ground truth data, which trains the neural networks. At runtime, the pre-trained networks provide robust BVH generation instantly without repeating the expensive construction process.
Data Source
AI summary
Techniques herein involve building bounding volume hierarchies for ray tracing using neural networks. These techniques use one trained neural network per animated mesh, with each such neural network being trained for a particular mesh topology. Meshes can be animated or otherwise modified to represent a single geometry object or portion of a geometry object in various animation states. Training a single neural network for each animated mesh allows such a neural network to generate BVHs for any animation state for the corresponding animated mesh in a robust manner. In other words, by limiting the responsibility of each such trained neural network to a single mesh topology (and therefore providing constraints to what the trained neural network must learn), it is possible for such a trained neural network to robustly and accurately generate BVHs.