Rendering with Neural Intersection Functions

Ray tracing involves computationally expensive calculations and geometry with varying degrees of complexity. According to Fujieda et. al[1], a novel Neural Intersection Function can be used for ray intersection queries. This method has better efficiency while ensuring image quality. The objective of this project is to implement and rigorously evaluate this Neural Intersection Function against the industry-standard bounding volume hierarchy (BVH) algorithm, providing a comprehensive comparative analysis of their respective performance characteristics.