MobileFace: 3D Face Reconstruction with Efficient CNN Regression

TL;DR

This paper introduces MobileFace, a compact and efficient CNN model capable of real-time 3D face reconstruction on mobile devices, achieved by leveraging morphable face models for training data and optimizing MobileNet architectures.

Contribution

The work presents a novel, fast CNN architecture for 3D face shape estimation that balances accuracy with real-time performance on mobile platforms.

Findings

Significant speed improvements over traditional methods.

Model size reduction enabling mobile deployment.

Maintains state-of-the-art accuracy in 3D face reconstruction.

Abstract

Estimation of facial shapes plays a central role for face transfer and animation. Accurate 3D face reconstruction, however, often deploys iterative and costly methods preventing real-time applications. In this work we design a compact and fast CNN model enabling real-time face reconstruction on mobile devices. For this purpose, we first study more traditional but slow morphable face models and use them to automatically annotate a large set of images for CNN training. We then investigate a class of efficient MobileNet CNNs and adapt such models for the task of shape regression. Our evaluation on three datasets demonstrates significant improvements in the speed and the size of our model while maintaining state-of-the-art reconstruction accuracy.