CVPR2021
Monocular Reconstruction of Neural Face Reflectance Fields
Mallikarjun B. R., Ayush Tewari, Tae-Hyun Oh, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Mohamed A. Elgharib, Christian Theobalt
摘要
Figure 1 . Given a monocular image, our method can synthesize One Light At a Time (OLAT) relit images, relight the face using any environment map (inset), and also transfer the light from another image (top) to the input (bottom). We can model view-dependent effects and can thus generate results in any head pose.