Lifespan Face Age Progression using 3D-Aware Generative Adversarial Networks

Today, face recognition systems (FRS) are widely used in applications such as forensic and border control systems. Despite the increasing ability of deep neural networks to identify individuals based on their facial images, it remains challenging to recognize faces with long age gaps between reference and probe samples. To improve the robustness of FRS towards aging, training datasets can be enriched with synthetic data by simulating recurring aging effects. Typical aging signs include craniofacial changes during the child-to-adult age transition or textural changes that occur during adult-to-adult aging (e.g., wrinkles or furrows). Building upon the recent achievements of 3D-aware generative adversarial networks, we propose a geometry-aware face age modification algorithm (Age-EG3D) that enables lifespan face age simulation. We demonstrate the effectiveness of our approach by providing a comprehensive performance evaluation and comparison of Age-EG3D to prior works. All code and pre-trained models are available at https://github.com/johndoe133/eg3d-age.