Supervisors
Luca Calatroni - DIBRIS
luca.calatroni@unige.it
About
Model-aware learning approaches for image reconstruction integrate the theoretical guarantees of model-based methods for inverse problems with the expressive capabilities of data-driven techniques. They are now widely used in imaging and frequently achieve state-of-the-art results when combined with appropriate optimisation schemes. Examples include optimisation-based methods (such as unrolling, plug-and-play) and physics-inspired generative approaches. When the underlying physics of the problem at hand is only partially known and/or highly complex and training data are limited, several difficulties may arise.
This project will focus on developing model-aware image reconstruction methods for partially unknown and potentially non-linear forward models in the presence of limited training data. To address data scarcity, we will employ self-supervised training procedures. Numerical validations will be performed on fluorescence 2D/3D+t microscopy data, in collaboration with the group of G. Vicidomini (IIT, Genova, Italy).
Expected starting date: May/June 25.
Duration & salary: The duration of the position is 2 years, and renewable. The salary varies depending on experience and is commensurate to international standards.
Perspective candidates should make an expression of interest before February 10, 2025.