COMPUTING - Devising advanced machine-learning methods for forensic f : Study

Project Description: Forensic facial reconstruction research is at the crossroads of art, medicine and science. It can be employed in the context of forensic investigation and for creating three-dimensional portraits of people from the past, from dead bodies and ancient Egyptian mummies to digital animations [1]. Previously, forensic anatomists and artists have relied on manually creating a 3D face clay model or through using computerized 3D forensic facial reconstruction software to manually reconstruct the face [2]. The aim of this multi-disciplinary PhD project is to develop advanced machine-learning and image-processing methods to learn how to automatically and accurately predict a facial'photograph' of a person from their skull (reconstructed using MRI), which can contribute to advancing both forensic medicine and science [2] as well as surface shape reconstruction from photos [3]. Developing such predictive models could be made in a fully quantitative manner with statistical uncertainty bounds on the predicted facial features as well as dynamic morphing of the predicted face within these bounds to aid recognition.