Reflectance and Natural Illumination from Single-Material Specular Objects Using Deep Learning.
Stamatios GeorgoulisKonstantinos RematasTobias RitschelEfstratios GavvesMario FritzLuc Van GoolTinne TuytelaarsPublished in: IEEE Trans. Pattern Anal. Mach. Intell. (2018)
Keyphrases
- deep learning
- multiple light sources
- single image
- light source
- specular reflection
- surface reflectance
- cast shadows
- point light source
- diffuse reflectance
- shape from shading
- reflectance properties
- dichromatic reflection model
- viewing conditions
- color constancy
- material properties
- real objects
- surface properties
- illumination direction
- unsupervised learning
- photometric stereo
- reflectance model
- d objects
- machine learning
- torrance sparrow
- image irradiance
- object surface
- illumination conditions
- weakly supervised
- surface normals
- multiple images
- lighting conditions
- spectral reflectance
- moving objects
- viewpoint
- segmentation method
- diffuse reflection
- reflection components
- partial occlusion
- textured surfaces
- mental models
- light transport
- surface roughness