AI to recognise emotions in faces, Gpi signs scientific article
Published in Frontiers in Artificial Intelligence, as part of the Parent project
As part of the Parent project, funded by the European Union to study the problems of newborn babies, a scientific article on the visual recognition of emotions has been published in “Frontiers in Artificial Intelligence”. Gpi is among the authors of the article with Angelo di Terlizzi, Vision and Innovation Manager of the Research and Development Department, and IT Researcher Syed Adil Hussain Shah.
Recognising emotions and diseases from faces
Face recognition is now a widely reached achievement of many different technologies, but the recognition of the emotions those faces express is still a frontier area. The next step is to identify and track disease progression through facial features, aiding in diagnosis and improving patient care, particularly in the presence of neurological disorders, cancers and autism.
Computer Vision and Deep Learning
The article discusses an innovative method that combines classical computer vision with Deep Learning for the extraction and identification of facial features, with a very high efficiency of 95 per cent, which is why the publication was accepted. The algorithm is called “Frai” and in the research described it was only applied on adults. However, a tuning phase is currently being concluded that will allow it to also be used to recognise pathologies in children, thus fitting within the scope of Parent.
Algorithm trained with health data
The algorithm will be trained using data provided by major health institutions: by learning from images of children who have developed pathologies, it will be possible to use the algorithm’s predictive capacity to detect the same problems in other subjects, helping the doctor to intervene in an early stage. It should be emphasised that the whole project is subject to the strictest rules on respecting people’s privacy.
Parent project, collaboration at European level
In the Parent project, Gpi is in a network with high-level organisations: for Italy Politecnico di Torino, Bambino Gesù Children’s Hospital and 7HC; for Spain Universidad de Cádiz and Fundación de la Investigación Biomédica; for Belgium KU Leuven and Icometrix; for Slovenia University of Ljubljana with NEUS Diagnostics; and finally for Switzerland Toelt Advance AI Lab.
