In this paper, we show how a 3D Morphable Model (i.e. a statistical model of the 3D shape of a class of objects such as faces) can be used to spatially transform input data as a module (a 3DMM-STN) within a convolutional neural network. This is an extension of the original spatial transformer network in that we are able to interpret and normalise 3D pose changes and self-occlusions. The trained localisation part of the network is independently useful since it learns to fit a 3D morphable model to a single image. We show that the localiser can be trained using only simple geometric loss functions on a relatively small dataset yet is able to perform robust normalisation on highly uncontrolled images including occlusion, self-occlusion and large pose changes.
@article{arxiv.1708.07199,
title = {3D Morphable Models as Spatial Transformer Networks},
author = {Anil Bas and Patrik Huber and William A. P. Smith and Muhammad Awais and Josef Kittler},
journal= {arXiv preprint arXiv:1708.07199},
year = {2018}
}
Comments
Accepted to ICCV 2017 2nd Workshop on Geometry Meets Deep Learning