[visionlist] Ph.D thesis offer on Deep learning of binary partition trees for image analysis

[lezoray]

Ph.D. Thesis proposal - University of Caen Normandy & GREYC CNRS Research Lab 

Title: Deep learning of binary partition trees for image analysis

Keywords : Hierarchical representations, Binary partition trees, Deep learning, Ultrametrics, Computer Aided Diagnosis of Skin lesions.

Subject
-------
There are many representations of digital images, each adapted to different contexts. In this thesis we are interested in hierarchical representations of images. These representations allow, from an over-segmentation of an image into super-pixels, to perform merging of regions at different scales. Such hierarchical representations allow to capture image features at different scales simultaneously, and are easily interpreted and manipulated by a human. Building good quality hierarchical representations is a very important step in image analysis. In image analysis, binary partition trees (BPT) are a popular hierarchical representation. Their construction relies on several key elements: an initial partition, a region model, a merging criterion, a merging order. This BPT construction often relies on region descriptors that are poorly suited to the data and on heuristic and greedy hierarchical clustering methods. We propose to take advantage of deep learning for the construction and manipulation of BPTs. The tree construction will then be able to exploit deep descriptors of superpixels, to learn the similarity between these descriptors and finally to have a learned merging criterion. As an ultrametric is a dual representation of a hierarchical representation, deep learning methods can be considered to learn not the ABP but directly the ultrametric from a graph representing the over-segmentation and by explicitly minimizing a cost function. The semantic segmentation of an image can then be seen as either a learned labeling of the vertices of the BPT or the learning of a cut in the BPT. A tree being a graph, convolution neural networks on graphs can be considered for this (convolution and pooling being very particular given the tree structure of the graph). Finally, applications in health (melanoma of the skin) and in satellite imagery will be made.

Qualifications
--------------
Candidates must have an MSc or engineering degree in a field related to computer science, electrical engineering, or applied mathematics, with strong programming skills (in particular with deep learning frameworks). Experience with image processing will be a plus. Candidates are expected to have abilities to write scientific reports and communicate research results at conferences in English.


Information and application: 
-----------------------------
Applications should include the following documents in electronic format: i) A short motivation letter stating why you are interested in this thesis, ii) A detailed CV describing your past research background related to the position iii)  The transcripts for master degrees. iv) The contact information for three references (do not include the reference letters with your applications as we will only ask for the reference letters for short-listed candidates). Please send your application package to olivier.lezoray at unicaen.fr and sebastien.bougleux at unicaen.fr.
The position will start in October 2023 and will be located in Caen, France. Ideally located in the heart of Normandy, two hours from Paris and just 10 minutes away from the beaches, Caen, William the Conqueror’s hometown, is a lively and dynamic city. 

Detailed pdf version
---------------------
Available at https://lezoray.users.greyc.fr/tmp/sujetTheseLezoray2023_en.pdf <https://lezoray.users.greyc.fr/tmp/sujetTheseLezoray2023_en.pdf>

Olivier LÉZORAY
Full Professor of Computer Science 

University of Caen Normandy
West Normandy Institute of Technology
Multimedia and Internet Department
F-50000 SAINT-LÔ+33(0)233775514 <tel:+33 2 33 77 55 14>	GREYC UMR CNRS 6072
Image Team - ENSICAEN
6 Bd. Marechal Juin
F-14000 CAEN+33(0)231452927 <tel:+33 2 31 45 29 27>
 <https://linkedin.com/in/olivier-lezoray-0983114/>	 <skype:olezoray>https://lezoray.users.greyc.fr <https://lezoray.users.greyc.fr/>
 <https://unicaen.fr/>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist_visionscience.com/attachments/20230216/bec370f0/attachment.html>