Real Time Neural Path Guiding
Bachelor thesis
A novel neural network based path guiding technique for realtime path tracing. A small neural network is trainied in online fashion to approximate incident radiance in world space. This information is used for guiding scatter rays in future frames.
Mar 30, 2023
Learning Surrogate Models for Fluid Dynamics
Seminar paper at KIT
Can the physical principles underlying expensive fluid dynamics simulations be approximated by neural networks? To answer this question, a literature review is conducted and various current research approaches are compared. Fundamental architectures such as CNNs, GNNs, and LSTMs are introduced, and their advantages and disadvantages are discussed based on the literature. The results show that neural networks are capable of approximating fluid simulations, often faster than traditional numerical methods.
Mar 15, 2021