we propose a novel framework to efficiently recover detailed facial geometry from calibrated multi-view images. The cost regression and multi-view fusion is solved by learning an implicit function in our network, which is proved to be more accurate and time-saving.
AAAI 2022 Paper (Oral)    Code