The backlog of unidentified missing persons is constantly growing. The ICRC’s requests to assist with disappearances continue to grow whilst the search activities remain heavy. As families get separated, they may contact the Red Cross Movement from varied locations, making it hard to liaise effectively between different data sources. Each caseworker may hold the lead to relevant information. These leads will then require further field tracing actions to clarify the fate and whereabouts of the missing person to provide their families with the answers they need and deserve. Forensic facial reconstruction by analysing an image of a skull can be used to determine the identity of human remains. This research aims to explore the possibility of using contemporary computer vision techniques, more specifically image-to-image translation models, to predict the 2D facial image of an individual from an image of their skull. This will by no means be considered a tool for identity confirmation but rather to generate leads to the possible identity of human remains. In addition, the proposed framework is designed to be low-cost and accessible (e.g. operating based on cell phone images taken by field workers without the need for advanced 3D scanners).
EPFL PI: Prof. Amir Zamir, Visual Intelligence and Learning Lab
Partner: Denise Abboud, ICRC