Assistant Professor, University of Rome - La Sapienza
Daniele Pannone received the M.Sc. and Ph.D. degrees in Computer Science from the Sapienza University of Rome, Rome, Italy, in 2015 and 2018 respectively. Since 2015, he is a member of the Computer Vision Laboratory (VisionLab) at the Department of Computer Science, Sapienza University of Rome, and since 2020 he is Assistant Professor at the same Department. His current research interests include scene analysis, pattern recognition, machine learning, deep learning, event recognition, object tracking, people reidentification, active vision in surveillance systems, signal analysis and processing, and human-computer interaction.
Daniele is a member of IAPR, CVPL and IEEE.
Road to AI: Computer Vision Tasks from Haralick Operators to Neural Networks
AI is becoming a predominant aspect of several Computer Science areas. However, most approaches can be considered as “black box” solutions where a given Neural Network (NN) model produces outputs without providing good insights into what happens inside the model itself. To address this aspect, a good starting point includes generating augmented inputs representing diverse aspects of the observed scene and extrapolating image characteristics, e.g., entropy feature map, to better understand what is being observed. While these procedures can be generated through renowned Computer Vision algorithms and offer the possibility to examine various facets of an image, they can also provide further insights into a NN architecture. In this context, the lecture follows the road to AI by introducing a powerful class of Computer Vision operators, i.e, Haralick’s textu.
An Application in Anomaly Detection