Inovações em reconstrução facial forense: análises de estimativa de sexo e afinidade populacional com inteligência artificial e machine learning

Abstract

Forensic Facial Reconstruction (FFR), also known as facial approximation, aims to estimate an individual's appearance by restoring lost soft tissues, using their skull as a reference. Although human identification relies on ante-mortem material for comparison, FFR serves as an alternative for image dissemination, assisting in recognition and the acquisition of comparative data. The techniques employed range from 2D to 3D approaches, either manual or computational, with population based averages of soft tissue thickness being one of the pillars of this process. However, the accuracy of reconstruction still faces challenges, particularly regarding the influence of the operator’s experience and the tools used. With technological advancements, new techniques have been developed to enhance the accuracy of digital facial reconstructions, including the use of force-feedback devices that allow for more precise virtual modelling. Nevertheless, questions remain about the reliability of methods used to estimate biological sex and population affinity, as well as the reproducibility of soft tissue marker placement, which directly impacts the reliability of the results. The incorporation of artificial intelligence and machine learning has the potential to minimize these variabilities, providing greater standardization and precision in forensic facial reconstruction.