Fernando Pérez González, Uvigo’s professor and director of the GPSC group from the atlanTTic center, gave one of the plenary lectures at the ICIAP 2017. A congress on image processing and analysis that is already in its 19th edition.
The congress was held on September 11th, 15th in Catania-Italy. It was attended by top-level researchers in different areas, among which we emphasized the application of processing technologies and analyzes to the forensic field, where Professor Pérez González has a long career.
His talk, entitled “Backstabbing Image Forensics”, was divided into two parts. The first one in which he set out in a synthetic way the state of the art and the past in the field of digital analysis of multimedia information, and a second part dealing with the present and trends in the field from a critical point of view, focusing on some statistical techniques and others that “abusing ” from machine learning.
From his point of view, the application of machine learning, although it is a technological trend (eg, neuronal networks) in many areas of our society, hides certain risks. One of them is the inability to know, once these networks are trained, what they really do; they are a black box. They are useful in many ways, but learning is not synonymous with understanding.
The application of this technological approach in forensic and judicial purposes still has some major implications. A false positive can help to send someone innocent to jail, in some cases without more technical explanation possible on the part of an expert than the legitimation of the technology used.
Fernando emphasized the risk of blind faith in this type of paradigm that, although useful, does not deepen into true knowledge as other more fundamental approaches, such as physics or mathematics, for example do.
More information on the congress: http://www.iciap2017.com/