We present a method for training a regression network from image pixels to 3D morphable model coordinates using only unlabeled photographs. The training loss is based on features from a facial recognition network, computed on-the-fly by rendering the predicted faces with a differentiable renderer. To make training from features feasible and avoid network fooling effects, we introduce three objectives: a batch distribution loss that encourages the output distribution to match the distribution of the morphable model, a loopback loss that ensures the network can correctly reinterpret its own output, and a multi-view identity loss that compares the features of the predicted 3D face and the input photograph from multiple viewing angles. We train a regression network using these objectives, a set of unlabeled photographs, and the morphable model itself, and demonstrate state-of-the-art results.

@inproceedings{Genova:2018:UTF,
   author = "Kyle Genova and Forrester Cole and Aaron Maschinot and Aaron Sarna and
      Daniel Vlasic and William T. Freeman",
   title = "Unsupervised Training for {3D} Morphable Model Regression",
   booktitle = "IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
      spotlight presentation",
   year = "2018",
   month = jun,
   pages = "8377--8386"
}