Réunion

Artificial Intelligence and Pattern Recognition in Remote Sensing

Annonce

As climate change intensifies and extreme events become more frequent, the monitoring and understanding of Earth system processes have become increasingly critical. Earth Observation (EO) and Remote Sensing (RS) support these efforts, driven by the rapid increase in the number of satellite missions. This growth in the volume and diversity of EO data has enabled large-scale environmental analyses, but has also introduced major challenges in data representation, interpretation, and timely analysis. In this context, advanced pattern recognition and data-driven techniques have become essential to extract meaningful information from these data. The objective of this meeting is therefore to review ongoing advances in EO and RS data analysis. To this end, we aim to cover the following topics during the day:

• Deep learning for Earth observation data
• Foundation models for Earth observation data
• Vision and language models for Earth observation
• 3D reconstruction
• Semantic classification and parameter estimation from remote sensing data
• Active, interactive and transfer learning
• Multi-modal and multi-temporal analysis
• Extraction, selection, learning, and reduction of features
• Novel pattern recognition tasks in remote sensing applications
• Explainable and interpretable machine learning
• Hybrid models, combining physics and machine learning
• Benchmark datasets

Invited speakers:

• Nicolas Audebert (IGN, LASTIG, STRUDEL)
  • Title: A tour of generative models for remote sensing
  • Abstract: Generative models are an untapped opportunity for remote sensing and Earth observation. They are versatile and unsupervised models that can produce an approximation of the data distribution. This makes them an effective tool for a large range of applications, from super-resolution to anomaly detection and domain adaptation. This talk will give an overview of modern classes of generative models, how they can be leveraged for Earth observation, and propose a roadmap for future research in « deep remote sensing ».
• Flora Weissgerber (ONERA, DTIS, SAPIA)
  • Title: Weakly supervised learning for the monitoring of the cryosphere
  • Abstract: The cryosphere changes rapidly under climate change forcing. On top of impacting hugely local communities, these changes drastically impact the future climate through feedback loops. In the Alps, both snow cover duration and depth are shortening, limiting access to fresh water or the capacity to produce hydroelectricity. At the poles, sea ice is becoming thinner and more fragile. This impacts wildlife and local sea-ice journeys, as well as increasing the albedo of the ocean and accelerating its warming. 
    Image processing technics, and in particular deep-learning, can expend the monitoring of these environments. Despite the large number of images available thanks to ambitious Earth observation programs, manual labels are generally very sparse. In this presentation, I will show how these sparse labels, different sensors, simulation and foundation models can be assembled through a good understanding of the physical properties of the imaged object, to design weakly supervised deep learning technics that overcome the limitations imposed by the label shortage. Firstly, I will present a weakly supervised deep learning algorithm to monitor seasonal snow combining both SAR and optical images. Then, I will present how sea ice can be mapped and its drift can be measured, combining SAR, optical images and nadir altimetry through dedicated weakly supervised deep learning algorithms.
    Down the line, we hope that these algorithms will help to improve the modeling of the future climate and that they could help local communities to adapt to climate change if they were transformed into accessible products.

Call for contributions :
We are welcoming contributions on these topics. We encourage presentations in English for international researchers, but do not restrict to it. Researchers and doctoral students wishing to present their work are invited to send their proposal (title and abstract) limited to one page by email by March 2, 2026, to:

• Charlotte Pelletier: charlotte.pelletier@univ-ubs.fr
• Sylvain Lobry: sylvain.lobry@u-paris.fr

Cette journée est labellisée par le comité technique 7 de l’IAPR

Organizers :

• Charlotte Pelletier (IRISA, Univ. Bretagne Sud)
• Ksenia Bittner (German Aerospace Center, DLR)
• Marc Russwurm (MEO-Lab, Univ. Bonn)
• Sylvain Lobry (LIPADE, Univ. Paris Cité)

Programme