The X-Faces Behind the Portraits of No One

This paper presents our work on the computational creation of photorealistic face images with a focus on how we transformed our generative and evolutionary system X-Faces into an interactive Media Art installation entitled Portraits of No One. The X-Faces system resorts to Computer Vision and Computer Graphics to automatically create new face images by recombining facial parts extracted from existing examples, along with Evolutionary Computation and Machine Learning to automatically explore the vast space of composite faces that can be generated. This system consistently generates face images with great photorealism and value, for example, for Data Augmentation or to assess and improve the performance of face detectors. In this paper, we describe how we explored the capabilities of X-Faces in a Media Art context. The result, the interactive installation Portraits of No One, synthesis and displays facial portraits on the borderline between the real and artificial using the facial features captured from its audience. The photorealism of the faces displayed in the installation space invoked the capabilities of Artificial Intelligence to generate content that makes people question its veracity. With the installation Portraits of No One, we allow people to engage with X-Faces and to be part of them.