Self-supervised Multi-level Face Model Learning for Monocular Reconstruction at over 250 Hz
This work addresses the need for fast and accurate monocular 3D face reconstruction, which is important for applications like virtual reality and facial analysis, and is incremental as it builds on existing parametric models with novel self-supervised learning.
The paper tackles the problem of reconstructing dense 3D face models from single images, which is challenging due to limited generalization of prior models, and presents a method that achieves state-of-the-art quality, better generalization to real-world faces, and runs at over 250 Hz.
The 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 state-of-the-art in terms of reconstruction quality, better generalizes to real world faces, and runs at over 250 Hz.