Lifting AutoEncoders: Unsupervised Learning of a Fully-Disentangled 3D Morphable Model Using Deep Non-Rigid Structure From Motion

Mihir Sahasrabudhe (1, 2) , Zhixin Shu (3) , Edward Bartrum (4, 5) , Riza Alp Güler (5) , Dimitris Samaras (3) , Iasonas Kokkinos (5)

1 CVN - Centre de vision numérique
2 OPIS - OPtimisation Imagerie et Santé
3 Department of Computer Science [Stonybrook - NY]
4 The Alan Turing Institute
5 UCL - University College of London [London]

In this work we introduce Lifting Autoencoders, a generative 3D surface-based model of object categories. We bring together ideas from non-rigid structure from motion, image formation, and morphable models to learn a controllable, geometric model of 3D categories in an entirely unsupervised manner from an unstructured set of images. We exploit the 3D geometric nature of our model and use normal information to disentangle appearance into illumination, shading, and albedo. We further use weak supervision to disentangle the non-rigid shape variability of human faces into identity and expression. We combine the 3D representation with a differentiable renderer to generate RGB images and append an adversarially trained refinement network to obtain sharp, photorealistic image reconstruction results. The learned generative model can be controlled in terms of interpretable geometry and appearance factors, allowing us to perform photorealistic image manipulation of identity, expression, 3D pose, and illumination properties.

Mots clés: deformable models; unsupervised learning; weakly supervised learning; deep learning; computer vision

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Apprentissage [cs.LG]; Traitement des images [eess.IV]; Intelligence artificielle [cs.AI]

Mihir Sahasrabudhe, Zhixin Shu, Edward Bartrum, Riza Alp Güler, Dimitris Samaras, et al.. Lifting AutoEncoders: Unsupervised Learning of a Fully-Disentangled 3D Morphable Model Using Deep Non-Rigid Structure From Motion. ICCV 2019 - IEEE International Conference on Computer Vision - Workshops, Oct 2019, Seoul, South Korea. ⟨hal-02422596⟩