Research (to be updated)
Phd
Title: Fusion de différents modes de capture pour la reconnaissance du visage appliquée aux e_transactions
Collaboration between Orange Labs. (Caen) and the GREYC (Image Team).
Location : Orange Labs., GREYC
Director: Pr. Marinette Revenu (GREYC-ENSICAEN)
Co-director: Dr. Olivier Lepetit
Approaches :
Convolutional Neural Networks (CNN) for face identification
Pretraining of CNN
Sparse representations of faces using learned dictionaries
Classifier fusion
CNN architecture for face reconstruction
Sparse decomposition of a face
Post-doc
Title : Segmentation et classification de cellules dans des grandes images en cytologie
Collaboration between the Image Team (GREYC), the CHPC (Centre Hospitalier Public du Cotentin) and the CHRU-Caen (Centre Hospitalier Régional Universitaire de Caen).
Location : GREYC (Image Team)
Length : from march 2011 to september 2011 (6 months)
Advisers : Pr. Olivier Lézoray, Pr. Abderrahim Elmoataz (GREYC)
Approaches :
Graph-based semi supervised segmentation for fast processing (images up to 100000x100000 pixels, containing hundreds of thousands cells)
(Multiscale) Deep networks for classification (see Feulgen database page)
Deep net for cells classification
Multiscale fusion
Post-doc
Title : Segmentation du foie dans des images SPECT/CT
Location : LITIS (University of Rouen), Quantif Team
Length : from november 2012 to october 2013
Advisers : Isabelle Gardin (Centre Henri Becquerel, Rouen), Pr. Su Ruan (LITIS)
Approaches :
Superpixels/supervoxels oversegmentation of the images/volumes (see ERGC page)
Graph-based semi supervised segmentation for fast processing
Software Organs Segmentation and Clustering for Application in Radiotherapy (OSCAR) actually evaluated at the Beaujon Hospital (AP-HP)
Initial slice
ERGC result (~1000 superpixels)
Manually placed seeds
Graph-based Segmentation
(~ 0.02 seconds)
Post-doc
Title : Inpainting de Vidéos stéréoscopiques
Location : GREYC (Image Team)
Length : from november 2013 to ...
Advisers : David Tschumperlé, Olivier Lézoray
Approaches :
Patch-based inpainting