Light Source Selection in Primary-Sample-Space Neural Photon Sampling
Event Type
Hybrid Formats
Registration Categories
TimeThursday, December 1610am - 6pm JST
LocationHall E (B2F, E Block)
DescriptionOur photon sampling is based a normalizing flow that is conditioned by a feature vector representing a light index. The conditioning allows us to represent a discrete probability distribution for light source selection in photon mapping, which is difficult by conventional neural importance sampling without the conditioning.