Please note: This PhD seminar will take place online.
Wenyou Wang, PhD candidate
David R. Cheriton School of Computer Science
Supervisor: Professor Toshiya Hachisuka
Traditional physically-based material models rely on analytically derived bidirectional reflectance distribution functions (BRDFs), typically by considering statistics of micro-primitives such as facets, flakes, or spheres, sometimes combined with multi-bounce interactions such as layering and multiple scattering. These derivations are often complex and model-specific. Once an analytic BRDF evaluation is defined, one still needs to design an importance sampling method for it and evaluate the probability density function (pdf) of that sampling distribution, requiring further model-specific derivations.
In this work, the authors present **PureSample**, a novel neural BRDF representation that allows learning a material’s appearance purely by sampling forward random walks on the microgeometry, which is usually straightforward to implement. Their representation enables efficient BRDF evaluation, importance sampling, and pdf evaluation for both homogeneous and spatially varying materials. This is achieved through two learnable components: first, the sampling distribution is modeled using a flow matching neural network, enabling both importance sampling and pdf evaluation; second, the authors introduce a view-dependent albedo term, captured by a lightweight neural network, which allows converting a pdf value into a BRDF value for any pair of view and light directions. The authors demonstrate PureSample on challenging materials, including various microgeometries, multi-layered materials, and multiple-scattering microfacet materials.