Seminar by Prof. Josiane Zerubia from INRIA, France
Prof. Josiane Zerubia from INRIA, France will give a talk entitled "Probabilistic Graphical Models and Deep Learning for Remote SensingImage Analysis" at CUHKSZ.
Abstract: Deep learning offers powerful non-parametric feature extraction, while probabilistic graphical models such as Markov random fields and conditional random fields provide a principled framework for structured prediction. This lecture examines their connections, from simple integration to end-to-end and intrinsically integrated formulations, and shows how they support remote sensing image analysis. Results on multimodal satellite and aerial imagery demonstrate applications in structured prediction, image fusion and segmentation, and image generation.
Speaker Bio: Professor Josiane Zerubia has been a permanent research scientist at INRIA since 1989 and Director of Research since 1995. She leads INRIA's AYANA exploratory research group, an interdisciplinary project spanning stochastic modeling, image processing, artificial intelligence, remote sensing and embedded computing. Her research focuses on image processing using probabilistic models, parameter estimation, statistical learning, optimization and artificial intelligence. She is an IEEE Life Fellow, EURASIP Fellow and IAPR Fellow, and an IEEE Geoscience and Remote Sensing Society Distinguished Lecturer (2025-2027).
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- 2001 Longxiang Blvd
- Shenzhen, Guangdong
- China
- Building: Chengdao Building
- Room Number: RM 207
Speakers
Probabilistic Graphical Models and Deep Learning for Remote Sensing Image Analysis
Deep learning offers powerful non-parametric feature extraction, while probabilistic graphical models such as Markov random fields and conditional random fields provide a principled framework for structured prediction. This lecture examines their connections, from simple integration to end-to-end and intrinsically integrated formulations, and shows how they support remote sensing image analysis. Results on multimodal satellite and aerial imagery demonstrate applications in structured prediction, image fusion and segmentation, and image generation.
Biography:
Professor Josiane Zerubia has been a permanent research scientist at INRIA since 1989 and Director of Research since 1995. She leads INRIA's AYANA exploratory research group, an interdisciplinary project spanning stochastic modeling, image processing, artificial intelligence, remote sensing and embedded computing. Her research focuses on image processing using probabilistic models, parameter estimation, statistical learning, optimization and artificial intelligence. She is an IEEE Life Fellow, EURASIP Fellow and IAPR Fellow, and an IEEE Geoscience and Remote Sensing Society Distinguished Lecturer (2025-2027).