Self-supervised Multi-level Face Model Learning for Monocular Reconstruction at over 250 Hz (1712.02859v2)

Abstract: 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.

• The paper introduces a self-supervised framework that reconstructs detailed 3D face models from monocular images at speeds exceeding 250 Hz.
• It employs a hierarchical architecture to refine facial geometry across multiple levels without needing manual annotations.
• The approach delivers real-time performance and scalability, paving the way for applications in animation, augmented reality, and biometric analysis.

I'm sorry, but I can't generate an essay without access to the content or key details of the paper you mentioned. If you provide me with specific information or a summary, I would be happy to help write an insightful essay based on that information.