Lorenzo Bruzzone
Speaking Sessions
Remote Sensing and the Labeled Data Challenge in the Foundation Model Era
Deep neural networks have become the primary approach for extracting semantic information from remote sensing data, supporting a wide range of tasks such as automatic classification, regression, and change detection. Despite the widespread use of supervised learning, a critical limitation in remote sensing applications is the scarcity of reliable labeled data for training deep models. This stands in contrast to the field of computer vision, where large-scale annotated datasets are more readily available. The limited availability of high-quality labeled samples highlights the need for weakly supervised learning strategies capable of effectively integrating multiple sources of imperfect labels into the training process. These weak labels are often associated with various types of uncertainty, which must be carefully modeled to enable robust learning.
This presentation addresses these challenges by offering a critical examination of learning in remote sensing, including the perspective offered by recent advancements in foundation models. It explores the diverse sources of labeled data and the inherent trade-offs between annotation quality and quantity. The main sources of uncertainty in weak labels—such as label noise, temporal mismatch due to land-cover changes, spatial ambiguity, and semantic inconsistency—will be analyzed in detail. Key approaches for robust weakly supervised learning will be discussed, including strategies for both single-date imagery and multi-temporal datasets. Methods such as few-shot learning, scribble-based supervision, and other forms of weak supervision will be explored, along with the potential and limitations of foundation models in reducing dependence on large-scale annotated datasets.
The presentation will illustrate these concepts through applications in climate change monitoring, precision agriculture, and planetary exploration.
Biography
Lorenzo Bruzzone received the Laurea (M.S.) degree in electronic engineering (summa cumlaude) and the Ph.D. degree in telecommunications from the University of Genoa, Italy, in 1993 and 1998, respectively. He is currently a Full Professor of telecommunications at the University of Trento, Italy, where he teaches remote sensing, radar, and digital communications.
Dr. Bruzzone is the founder and the director of the Remote Sensing Laboratory (https://rslab.disi.unitn.it/) in the Department of Information Engineering and Computer Science, University of Trento. His current research interests are in the areas of remote sensing, radar and SAR, signal processing, machine learning and pattern recognition. He promotes and supervises research on these topics within the frameworks of many national and international projects. He is the Principal Investigator of many research projects. Among the others, he is currently the Principal Investigator of the Radar for Icy Moon Exploration (RIME) and of the Subsurface Radar Sounder (SRS) instruments in the framework of the JUpiter ICy moons Explorer (JUICE) and the Envision missions, respectively, of the European Space Agency (ESA) and is the for the High Resolution Land Cover project in the framework of the Climate Change Initiative of ESA. He is the author (or coauthor) of more than 400 scientific publications in referred international journals, more than 390 papers in conference proceedings, and 25 book chapters. He is editor/co-editor of 18 books/conference proceedings and 2 scientific books. His papers are highly cited, as proven from the total number of citations (more than 53400) and the value of the h-index (106) (source: Google Scholar). He was invited as keynote speaker in more than 40 international conferences and workshops. He was a member of the Administrative Committee of the IEEE Geoscience and Remote Sensing Society (GRSS) between 2009 and 2023. Dr. Bruzzone ranked first place in the Student Prize Paper Competition of the 1998 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Seattle, July 1998. Since that he was recipient of many international and national honors and awards, including among the most recent ones the IEEE GRSS 2015 Outstanding Service Award, the 2017 and 2018 IEEE IGARSS Symposium Prize Paper Awards, the 2019 WHISPER Outstanding Paper Award and the 2022 Letter Prize Paper Award for the best paper published on the IEEE Geoscience and Remote Sensing Letters in 2022. Dr. Bruzzone was a Guest Co-Editor of many Special Issues of international journals. He is the co-founder of the IEEE International Workshop on the Analysis of Multi-Temporal Remote-Sensing Images (MultiTemp) series and is currently a member of the Permanent Steering Committee of this series of workshops. Since 2003 he has been the Chair of the SPIE Conference on Image and Signal Processing for Remote Sensing. He has been the founder of the IEEE Geoscience and Remote Sensing Magazine for which he has been Editor-in-Chief between 2013-2017.Currently he is an Associate Editor for the IEEE Transactions on Geoscience and Remote Sensing. He has been Distinguished Speaker of the IEEE Geoscience and Remote Sensing Society between 2012-2016. He is a Fellow of IEEE.