We present a novel face reconstruction method capable of reconstructing detailed face geometry, spatially varying face reflectance from a single monocular image. We build our work upon the recent advances of DNN-based auto-encoders with differentiable ray tracing image formation, trained in self-supervised manner. While providing the advantage of learning-based approaches and real-time reconstruction, the latter methods lacked fidelity. In this work, we achieve, for the first time, high fidelity face reconstruction using self-supervised learning only. Our novel coarse-to-fine deep architecture allows us to solve the challenging problem of decoupling face reflectance from geometry using a single image, at high computational speed. Compared to state-of-the-art methods, our method achieves more visually appealing reconstruction.
@article{arxiv.2203.07732,
title = {S2F2: Self-Supervised High Fidelity Face Reconstruction from Monocular Image},
author = {Abdallah Dib and Junghyun Ahn and Cedric Thebault and Philippe-Henri Gosselin and Louis Chevallier},
journal= {arXiv preprint arXiv:2203.07732},
year = {2022}
}