Efficient Spatial Hierarchies for Complex 3D Scenes

The project focuses on spatial data structures with an emphasis on bounding box hierarchies (BVH). Our main research goal is to develop new methods for constructing and traversing BVHs that improve their efficiency and scalability. We focus on using new tight bounding boxes in combination with efficient parallel algorithms, an area that is still underexplored. We also focus on integrating deep learning techniques to solve selected subtasks within BVH methods and designing automated scene preprocessing methods that better utilize the existing BVH support in modern graphics APIs.