Missions

The context of this postdoctoral fellowship is to contribute to the introduction of new hybrid methods, in the sense that they bridge the gap between "classical" methods (energy minimization, partial differential equations, etc.) and recent Deep Learning-type methods. The focus will be on geodesic methods, the search for paths of minimum length according to an adapted metric, imposing for example a penalization of the curvature. In addition, this research is motivated by applications in medical imaging, such as the precise segmentation of regions of an image.

Activités

This work will take place at CEREMADE, Université Paris Dauphine PSL, in Paris 16th.
It will include the introduction of new methods, their implementation, development,
validation, as well as the writing of conference and journal articles.
It may include co-supervision of students.

Compétences

PhD in image analysis, applied mathematics or computer science.
Strong knowledge in geodesic methods and Deep Learning.
Strong programming skills.
Motivation for applications in medical imaging.
Publications in recognized journals in the field (IEEE TPAMI, IJCV, JMIV, IEEE TMI, IEEE TIP, …)
and/or major conferences (ICCV, CVPR, …)
Proficiency in written and spoken English.

Contexte de travail

The recruited person will join CEREMADE, and will be hired by the National Center for Scientific Research (CNRS) as part of a funding by the ANR. The contract is for one year.

Contraintes et risques

N/A