Material and Colored Illumination Separation from Single Real Image via Semi-Supervised Domain Adaptation
SessionPosters Gallery
DescriptionWe propose a training strategy for intrinsic decomposition networks that bridges the domain gap between synthetic and real images, enabling even simple CNN to achieve excellent material and illumination separation.
Contributors
Event Type
Poster
TimeThursday, 5 December 20248:00am - 6:00pm JST
LocationLobby Gallery (1) & (2), G Block, Level B1
TE
EH
