Thematic Session 2

THEMATIC SESSION #02

Filippo Maggioli

Pegaso University

• filippo.maggioli@unipegaso.it

Andrea Generosi

Pegaso University

• andrea.generosi@unipegaso.it

Valerio De Luca

Pegaso University

• valerio.deluca@unipegaso.it

Diana Marin

Technische Universität Wien

• diana.marin@tuwien.ac.at

This session explores methods, models, and applications for automatic understanding of scenes observed from a first person perspective, as they are perceived by the user. In such a context, various kinds of data received from wearable devices (such as smart glasses and XR headsets) can be exploited, with the field of view aligned with the user's gaze and actions.

The session aims at creating a meeting point between computer vision, human-centered AI, and eXtended Reality, promoting scalable solutions that could be used in real-world, uncontrolled scenarios, characterised by high environmental variability, unpredictable lighting, occlusions, sensor noise, and spontaneous user behaviour. The session focuses on challenges related to scene dynamics, user-environment interaction, model generalisation, and robustness of algorithms outside laboratory environments.

Theoretical, methodological and applied contributions are encouraged, including (but not limited to) the following topics:

• Egocentric vision and first-person scene understanding
• Recognition of actions, activities and intentions from an egocentric perspective
• Understanding hand-object interactions and affordance
• Integration of multimodal signals (video, depth, gaze, IMU, audio)
• Gaze tracking, attention modeling and context awareness
• Robust learning and generalisation in uncontrolled environments
• Egocentric datasets “in the wild” and evaluation protocols
• Self-supervised and continual learning for egocentric data
• XR applications for contextual assistance, industrial training and healthcare
• Implications for privacy, ethics and reliability of egocentric systems

Filippo Maggioli is currently a Tenured Assistant Professor at Pegaso University, Department of Information Science and Technology. Previously, he was Adjunct Professor at Pegaso University, Postdoctoral Researcher at the University of Milano-Bicocca, and Research Intern at the King Abdullah University of Science and Technology. Filippo defended his Ph.D. thesis on “Scalable Geometry Processing for Computer Graphics Applications” in 2023 at Sapienza – University of Rome. His research activity is in the fields of Computer Graphics and Computer Vision, with a particular focus on Geometry Processing, Computational Geometry, Shape Analysis, and Spectral Geometry. Other research interests also span the fields of Physical Simulations and Parallel Computing.

Andrea Generosi is a Tenure-Track Assistant Professor at the Department of Information Science and Technology, Pegaso University (Italy). His research focuses on the design and development of innovative solutions based on deep learning. PhD in Industrial Engineering, engaged in research and development projects for industry and the cultural sector, with a focus on Affective Computing, ergonomics, automotive, HCI and Extended Reality. Previously a Research Fellow and adjunct professor at the Polytechnic University of Marche and co-founder of the UNIVPM Emoj Srl startup and spin-off.

Valerio De Luca has been an Assistant Professor at the Department of Information Science and Technology, Pegaso Telematic University (Italy) since December 2024. He has been collaborating with the Augmented and Virtual Reality Laboratory (AVR Lab) of the University of Salento since 2015, working on projects in human–computer interaction and extended reality applied to medicine, education, and cultural heritage. His recent research focuses on extended reality for preoperative planning and intraoperative support in surgery, integrating computer vision techniques based on stereo cameras with inertial sensor data. His previous work includes GPU computing for UAV path planning, augmented reality for improving situation awareness in UAV control, and research on grid computing, distributed multimedia systems, and Quality of Experience (QoE).

Diana Marin is a postdoctoral researcher in the Virtual and Augmented Reality Group at TU Wien, led by Prof. Hannes Kaufmann. She earned her PhD in February 2025 on proximity-based point cloud reconstruction in the Rendering and Modeling Group at TU Wien under the supervision of Prof. Michael Wimmer and Dr. Stefan Ohrhallinger. Her research combines geometry, topology, and human perception to improve how unstructured data is reconstructed, interpreted, and interacted with, with applications including clustering, segmentation, and procedural generation.