INTERSHIP Implicit Neural Representation for Lake & River Floodplain Topography

The TPA service of the Data Campus has developed a 3D representation system for the state evolution of lakes/reservoirs and river reaches using SWOT time series. This includes estimating 3D bathymetry, water fill level, water surface area, volume changes, and their associated uncertainties by leveraging implicit neural representation techniques.

This type of representation accounts for measurement uncertainties at both the input and output of the model, as well as the efficient large-scale processing of SWOT time series.

The objective of the internship will be to integrate the contributions of new satellite sensors and data sources into the implicit network: - SWOT KaRIn interferometric SAR images (Single Look Complex) and/or derived point clouds - Sentinel-1 SAR images (SLC and/or GRD) - Sentinel-2 L2A (optical imagery) - Sentinel-3 / Sentinel-6 / SWOT Nadir (Doppler altimetry) The envisioned approaches may build upon existing literature in image and signal processing, including fields outside of remote sensing such as Bayesian methods, deformable models, random fields, and optimal transport - Final-year student in an engineering school or Master's degree, specializing in computer science, applied mathematics, or image and signal processing. - Strong knowledge of image processing, applied mathematics, and programming (preferably Python) is required. - Skills in remote sensing and the ability to develop in a Linux environment would be an asset. - Prior knowledge of hydrology is not required. - The candidate should demonstrate strong organizational skills in their daily work.