General information
Reference : UPR8001-SIMLAC-012
Nombre de Postes : 1
Workplace : TOULOUSE
Date of publication : Monday, January 23, 2023
Scientific Responsible name : Simon Lacroix
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 15 March 2023
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly
Description of the thesis topic
LAAS has designed a hyperspectral imaging system that allows the selection of information from a scene (compressed acquisitions) thanks to a controllable array of micro-mirrors. This control capability allows to design acquisition schemes adapted to the observed scene and to the observation objectives.
The purpose of this thesis is to integrate compressed data classification techniques with decision techniques to define optimal acquisition sequences. For these two problems, two learning approaches will be exploited: convolutional neural networks for classification, which exploit in particular the direct acquisition model of the compressed data, and deep reinforcement learning for the definition of complementary acquisitions.
The PhD will be supervised by Simon Lacroix (LAAS), and Hervé Carfantan (IRAP)
Work Context
The thesis will be co-supervised by Hervé Carfantan, from the team Signals and Images for the Sciences of the Universe of the IRAP in Toulouse, with which the LAAS has been collaborating closely for a few years on the processing of the data acquired by the prototype imager considered.
We talk about it on Twitter!