KEYNOTE
Martian Images Segmentation and Classification through Green AI
Francesco Carlo Morabito
AI_Lab, University "Mediterranea" of Reggio Calabria, Italy
ABSTRACT
Artificial Intelligence (AI) approaches based on Machine and Deep Learning (ML/DL) are rapidly emerging in varied fields as a paradigm shift in solving complex problems. They are basically founded on huge data availability representing the phenomenon/process under analysis. From the data, that are supposed to include most of the relevant modes of the process, it is possible to extract pattern representative of what we are searching for. Once deployed on suitable hardware, this representation, alternative to the traditional model-based one, can be used to generalize on fresh data, potentially in real-time. However, the use of AI requires a systematic design in order to ensure efficiency, quality and reliability: this is particularly relevant in space applications (i.e. critical missions), where we need high performance coupled with trustworthiness and interpretability.
Recently, NASA’s Perseverance Mars rover explored the surface of Mars by autonomously analyzing data to seek out and discover specific minerals in the rocks. It took autonomous decisions based on real-time analysis of rocks’ composition.
In this talk, I will present a different application that aims to extract important information on the presence of specific objects in martian images. AI is trained on synthetic, augmented and real martian images to classify terrain characteristics (i.e., soil, rocks, sands, …), by considering varied lighting conditions, inclination, texture and colors. Some Curiosity rover images are also analysed. The approaches here proposed are also designed to be energy-aware (Green-AI).
The presented methodological approaches and results are related to the project Standardizing Trustworthiness and Neural Development for AI (Stand4AI), funded by ASI, which we are developing in partnership with Thales Alenia Space Italia. The main objective here is to develop AI systems that can be space qualified in particular through the novel concept of explainability (xAI).
SPEAKER BIOGRAPHY
Francesco Carlo Morabito is a Full Professor (2001) of Electrical Engineering and Neural Engineering with the University “Mediterranea” of Reggio Calabria, Italy. He served there as Dean (Faculty of Engineering, 2001-2008), as President of the Courses in Electronic Engineering (1998) and Industrial Engineering (2015), as Vice-Rector for Internationalisation (2013-2022), and as Deputy Rector (2017-2018). He’s the founding Director of NeuroLab and AI_Lab, at DICEAM, UNIRC. He has authored or co-authored over 400 papers in international journals/conference proceedings in various fields of engineering (machine/deep learning, biomedical signal processing, radar data processing, nuclear fusion, nondestructive evaluation, artificial and computational intelligence). He has co-authored >25 intl. books (mostly focused on neural networks and machine learning) and held five international patents. Prof. Morabito is a Foreign Member for the Royal Academy of Doctors, Spain (2004-), a Foreign Member for the Royal Academy of Economic and Financial Sciences (RACEF, 2025-), a member of the Institute of Spain, Barcelona Economic Network (2017-). Senior Member of IEEE (2000), Life senior Member (2025-), and of INNS (2006). Governor of the International Neural Network Society (INNS), 2022-2024, and earlier for 12 years (2000-2012). President of INNS (2024-2026). He served as President of the Italian Network Network Society (SIREN), 2008-2014, and is co-chair of the Italian Conference on Neural Networks (WIRN). Editorial Board member for Neural Networks, International Journal of Neural Systems, and EiC for Artificial Intelligence in Neurology. He’s included since 2021 in the Top 2% researchers according to the University of Stanford/Elsevier database.