Deep learning for earth observation data (with Danish Hydrological Institute)

Created on August 19, 2026

2026   ·   Beyond medical imaging

Supervisors: Veronika Cheplygina (ITU), Spyros Kondylatos (DHI)

Earth observation (EO) has become a fundamental tool for monitoring environmental processes, land use dynamics, and water systems at regional and global scales. Satellites such as the Sentinel-1 and Sentinel-2 provide high-quality data with global coverage and frequent revisit times. Despite their strengths, there are several challenges which need to be addressed for trustworthy and reliable deep learning models, such as:

  • Dataset creation with fine-grained annotation about objects and structures visible at ground level
  • Uncertainty estimation of deep learning models
  • Strategies (e.g. dataset distillation, transfer learning) for scenarios with limited labeled data

For more details about the projects, please see the DHI website: https://dhi.github.io/student-project-catalogue/

Multiple projects are available, suitable for students with experience with machine learning and HPC. Groups of 2 preferred.