Invited Speakers

Title: Deep Learning for Image and Video Forensics: Mine of Information or Minefield?

Presenter: Martino Jeriam, Amped Software, CEO and Founder

 Abstract

Deep Learning is revolutionizing many fields and changing our society. Its impact on image processing and analysis cannot be overstated, however, there are some applications - such as those related to criminal investigations and justice - that must be handled with care.

In this presentation, we will discuss the possible pitfalls of using deep learning for forensic applications. To do so, we distinguish between the evidentiary and investigative use of image processing, highlighting the difference in requirements between the two cases. We also separate the case of image enhancement and image analysis, including authentication, and we provide our view on the suitability of deep learning-based image processing in each scenario.

In addition to this, we will see a few applicative examples studied and published at Amped Software in relation to facial image enhancement, low-quality license plate detection, and deepfake analysis.

Presenter Bio

Martino Jerian is the CEO and Founder of Amped Software. He graduated in Electronic Engineering at the University of Trieste, Italy (summa cum laude) with a thesis on forensic image processing and founded Amped Software in 2008. He has extensive software engineering experience, having designed and started the development of Amped Software products. He taught as a contract professor in a few university courses related to investigations, forensics, and intelligence. He has also published several scientific papers related to image and video forensics and worked as a forensic expert in major judiciary cases. His current activities are mainly focused on the improvement of the relationship between science and justice during forensic investigations, the need for proper education and training on image and video analysis, and raising awareness on the challenges related to the proper use of video evidence, especially in relation to cybersecurity and AI.