A Deep Deformable Convolutional Method for Age-Invariant Face Recognition

Abstract

With the rapid development of deep learning, face recognition also finds its improving dramatically. However, facial change is still a main effect to the accuracy of recognition, as some complex factors like age-invariant, health state and emotion, are hard to model. Unlike some previous methods decomposing facial features into age-related and identity-related parts, we propose an innovative end-to-end method that introduces a deformable convolution into a deep learning discriminant model and automatically learns how the facial characteristics changes over time, and test its effectiveness on multiple data sets.