The developed Applications include:

• Artificial Intelligence (AI) for Airport Operations: the aim is to develop a digital twin of an airport that uses computer vision modules to collect real-time data on aircraft, vehicles and operators during ground operations. This data will be integrated and visualized in various ways to improve situational awareness and support efficient operations, especially in extreme weather events. The system includes modules for video processing, data collection and integration, and visualization through maps and augmented reality, as well as a simulator to train operators in a virtual environment. The goal is to improve the resilience of the airport and decision-making processes.
• Generative AI techniques for virtual reconstruction of frescoes: these techniques, emerged in recent years, can automatically create multimedia content of any type. In this specific case, they have been applied in the field of digital humanities to create images, that show a possible solution for the virtual reconstruction of damaged, or with missing portions, frescoes of the Castle of Mirabello. Goal was to create immersive and interactive experiences for a virtual visit of the castle, with the integration of real and virtual items into a digital context.

Furthermore, in the past:

• Automatic visual identification of damage in buildings and infrastructure: deep Learning methods for structural damage detection, including object detection, open-vocabulary detection, visual language models, segmentation.
• Automatic detection techniques for accidental falls: in 2018-2021 we have used smart wearable devices and deep learning on embedded to detect accidental falls with recurrent neural networks. Datasets with simulated falls by volunteers have been collected: seven carry positions, seventeen different activities, forty volunteers, over five thousands tracks. Manual annotations on videos, basic for training, have been done.
• Deep reinforcement learning for collaborative robotics: focused on incremental autonomous learning, experience transfer, and robust avoidance strategy, with the virtualization of a real‐world robot to learn reaching a target while avoiding obstacles in a simulation environment.

---

Latest publications

• Marco Piastra and Patrizia Catellani (2025). On the Emergent Capabilities of ChatGPT 4 to Estimate Personality Traits, in Frontiers in Artificial Intelligence, Sec. AI for Human Learning and Behavior Change, Volume 8, DOI:10.3389/frai.2025.1484260, part of the Research Topic Human-Centered Artificial Intelligence in Interaction Processes.
• Awais Khan Nawabi, Janos Tolgyesi, Elena Bianchi, Chiara Toffanin, Piercarlo Dondi (2025). D-Care: A Multi-Tone LLM-based Chatbot Assistant for Diabetes Patients, in Proceedings of the 18^th International Joint Conference on Biomedical Engineering Systems and Technologies - HEALTHINF, SciTePress, pp. 766-773, ISBN 978-989-758-731-3, ISSN 2184-4305, DOI:10.5220/0013266600003911.
• Angela Rosy Morgillo, Stefano Mangini, Marco Piastra, Chiara Macchiavello (2024). Quantum state reconstruction in a noisy environment via deep learning, Quantum Machine Intelligence, 6, 39, DOI:10.1007/s42484-024-00168-x.
• Alessandro De Carlo, Davide Ronchi, Marco Piastra, Elena Maria Tosca, Paolo Magni (2024). Predicting ADMET Properties from Molecule SMILE: A Bottom-Up Approach Using Attention-Based Graph Neural Networks, Pharmaceutics, 16(6):776, DOI: 10.3390/pharmaceutics16060776.
• Patrizia Catellani, Marco Biella, Valentina Carfora, Antonio Nardone, Luca Brischigiaro, Marina Rita Manera, and Marco Piastra (2023). A theory-based and data-driven approach to promoting physical activity through message-based interventions, in Frontiers in Psychology, 14, 1200304.