Primo Zingaretti
Full Professor at Università Politecnica delle Marche
Primo Zingaretti is Full Professor of Computer Engineering at the Università Politecnica delle Marche. His main research interests have been in the areas of artificial intelligence (machine and deep learning, decision support systems), (mobile) robotics, intelligent mechatronic systems, computer vision, pattern recognition, image processing and understanding, information systems, learning, e-health and e- government. Robotics vision (aerial, ground and underwater autonomous systems), mechatronic/embedded/cyber physical systems, human behaviour analysis (in retail and home automation/ambient assisted living) and geographical information systems (remote sensing from satellite and/or UAV, GIS-ready automatic cartography, precision farming) have been the main application areas. He was a founding partner of two university spin-offs and a scientific and/or technical responsible for many national and international research projects, funded by private companies or by public bodies. He has been General Chair or Program Chair of international conferences and he is author or co‐author of more than 200 scientific papers in journals, book chapters or conference proceedings, Senior Member of IEEE, promoter of AIxIA, Vice President of CVPL.
Lecture:
Computer Vision and Deep Learning for Human Behavior Analysis and Human Robot Interaction
Abstract:
Following the transition to the era of multimedia big data, machine learning approaches have evolved into deep learning approaches, leading to a more powerful and efficient way of tackling complex problems. This talk will illustrate how understanding consumer behaviour and, more generally, human behaviour analysis (HBA), long studied using feature-based computer vision techniques and machine learning concepts, has now been replaced by deep learning approaches. Some end-to-end deep learning solutions (e.g. for people counting, classification of user interactions and re-identification of people) in intelligent retail environments, where understanding consumer behaviour is of great importance and one of the keys to success for retailers, will be presented. Some (retail) datasets for DL approaches, trajectory analysis and prediction (in indoor and outdoor, real and virtual environments), the use of generative modelling for trajectory generation, “senseable” spaces (a term coined to define the types of spaces that can provide users with contextual services, measure and analyse their dynamics and react accordingly, in a seamless exchange of information), and some human-robot interaction tasks will also be presented.
