279 Foundation Models for Multimodal Satellite Imagery Data Processing

Job title:

279 Foundation Models for Multimodal Satellite Imagery Data Processing

Company:

Centre National d’Etudes Spatiales

Job description

Your application must include a recommendation letter from your Ph.D. supervisor, a detailed CV including university education and work experience, a list of publications, a 2-page description of the work undertaken during the course of your PhD.For more Information, contact : Directeur de RechercheSubmit the complete application online (Apply) before March 14th, 2025 Midnight Paris timeThis postdoctoral research project aims to explore the application of foundation models (FMs) in the processing and analysis of multimodal satellite imagery data. The advent of foundation models is a major breakthrough in various fields, including Earth Observation ( ). These models, pre-trained on vast amounts of data, have shown great potential in improving the accuracy and efficiency of satellite imagery analysis. By exploring the potential of FMs to extract meaningful information from diverse satellite data sources, this research seeks to address technological locks and challenges raised by multimodality for Earth observation uses such as environmental monitoring, urban planning, and disaster response.Multimodality in satellite imagery refers to the integration of various data types from different sources, such as optical images, radar, thermal data. While this integration has significant potential for enhancing analyses and applications, it poses several challenges:– Dataset Challenges: Data skew, Data availability, Data quality challenges. All of this can lead to an unbalanced dataset, certain modalities might be overrepresented, with the risk to obtain biased models.– Integration of Heterogeneous Data ( ), different spatial and temporal resolutions, diverse feature spaces coming from different types of information captured by each modality), noise and artifacts, different ground projections (geometric models, DTMs)To address these challenges, initial work could focus on the characteristics of available large national EO datasets (e.g. Spot World Heritage) and on the following areas:

• Data Preprocessing and Fusion: Data alignment, registration and normalization

• Feature Extraction and Representation: feature selection, managing redundancy and complementarity, extract and use feature relationships

• Model Fusion and Integration: early fusion and concatenation or intermediate fusion guided by features selection

ProfilPhD in computer science (or environment/Earth observation) with a specialization in data analysis, statistics, AI.

Expected salary

Location

Toulouse

Job date

Wed, 05 Feb 2025 01:16:38 GMT

To help us track our recruitment effort, please indicate in your email/cover letter where (vacanciesin.eu) you saw this job posting.