Our novel monocular reconstruction approach estimates high-quality facial geometry, skin reflectance (including facial hair) and incident illumination at over 250 Hz. A trainable multi-level face representation is learned jointly with the feed forward inverse rendering network. End-to-end training is based on a self-supervised loss that requires no dense ground truth.
AbstractThe reconstruction of dense 3D models of face geometry and appearance from a single image is highly challenging and ill-posed. To constrain the problem, many approaches rely on strong priors, such as parametric face models learned from limited 3D scan data. However, prior models restrict generalization of the true diversity in facial geometry, skin reflectance and illumination. To alleviate this problem, we present the first approach that jointly learns 1) a regressor for face shape, expression, reflectance and illumination on the basis of 2) a concurrently learned parametric face model. Our multi-level face model combines the advantage of 3D Morphable Models for regularization with the out-of-space generalization of a learned corrective space. We train end-to-end on in-the-wild images without dense annotations by fusing a convolutional encoder with a differentiable expert-designed renderer and a self-supervised training loss, both defined at multiple detail levels. Our approach compares favorably to the stateof-the-art in terms of reconstruction quality, better generalizes to real world faces, and runs at over 250 Hz.