Pose and Expression Robust Age Estimation via 3D Face Reconstruction from a Single Image

In this paper, we present a deep learning architecture that exploits 3D face reconstruction to obtain a robust age estimation. To this end, effective representation is learned through an expression-, pose-, illumination-, reflectance-, and geometry-aware deep model reconstructing a 3D face from a single 2D image. The 3D face reconstruction network is combined with an appearance-based age estimation network, where the face reconstruction features are jointly learned with the visual ones. Experiments on large-scale datasets show that our method obtains promising results and outperforms state-of-the-art methods, especially in the presence of strong expressions and large pose variations. Furthermore, cross-dataset experiments show that the proposed method is able to generalize more effectively as opposed to the state-of-the-art methods.