# General Information 3-Year PhD position in Saint-Etienne on « Enrichment models for temporal heterogeneous image data – Application to the analysis of micro-expressions ». Starting date: September or October 2015 Application deadline date: 15th of April Decision announcement date: june 4th 2015 # Context The images have an increasing place in our society. New technologies are made available to the public (smartphones, kinect, Head-Up Displays (HUD), ...) bringing new requirements or changes in behavior. We must support this change which emanates the need to capitalize plethora of information contextually exploited and with constraints. Hubert Curien laboratory has defined among its priorities, the study of new structuring models of visual information closely linked with learning methods. These models should be able to integrate information of different types (color, depth, ...) according to their spatial and temporal dimensions to address problems that may appear in different areas of imaging application. # PhD thesis TThe objective of this PhD thesis is to develop one (or more) formalism(s) for addressing heterogeneous image data that change over time. Thus, from a video, modeling of visual events can be mainly done using color, textures, contours and movement. From a sequence of depth images, it is possible to use depth, of course, but also local normals of the surfaces and their evolution in time. For thermal or spectral image sequences, we will also find this duality between spatial information and directional information. The proposed formalisms can exploit probabilistic modeling [CS12 ], the geometry of information [ha14] nonparametric Bayesian methods [EV14], among others. TThe chosen application will be the analysis of micro-expressions, that would better characterize the states such as stress, fatigue, pain, ... We will focus on the characterization of pain, which is a recent area of ​sstudy under development [FERA15] [WG13]. The proposed methods will then describe micro-expressions (squinting, motion of wrinkles, etc.) that can appear in different image modalities (color, depth, thermal imaging, etc.) in order to identify pain levels. To this aim, the following questions related to the definition of a mutual and contextualized enrichment of data will arise: how to properly manage the complementary information; how to achieve detect rare but sometimes very important events for a meaningful analysis of information; how to successfully manage the links between the data for the modeling of an event and its evolution in time. To validate our work and compare our experimental results with recent proposed methods in the field, we plan initially to use existing databases such as the one described in [WG13]. If necessary, we will implement an experimental protocol to develop an appropriate database. # Student SWe are looking for a student with a Master obtained at least the 1st of September 2015, in the field of "computer science" with a good background in applied mathematics (probability, data analysis, estimation and optimization, ...) and software development (algorithmic, project implementation in C / C ++ and / or Matlab - Octave - Scilab and / or Python ...). Knowledges in image processing, RGB-D data processing, and / or machine learning would be appreciated. # Salary Net salary: 1367 euros without teaching activities and 1643 euros with teaching activities (64 hours teaching each year). # Application Your application should include the following documents: - Letter of intent - Grades and ranking during Master 1 and Master 2 - Scientific CV - List of publications (if it exists of course) - Referees Contacts : - olivier.alata@univ-st-etienne.fr (http://perso.univ-st-etienne.fr/ao29170h/) - anne.claire.legrand@univ-st-etienne.fr - hubert.konik@univ-st-etienne.fr - remi.emonet@univ-st-etienne.fr (http://home.heeere.com/) # Bibliography [CS12] A. Carmi, F. Septier and S. J. Godsill., « The Gaussian mixture MCMC particle algorithm for dynamic cluster tracking », Automatica, 48 (2012) 2454–2467. [EV14] R. Emonet, J. Varadarajan, J.M. Odobez, « Temporal Analysis of Motif Mixtures using Dirichlet Processes », TPAMI (Transactions on Pattern Analysis and Machine Intelligence), 2014. [FERA15] Challenge Fera2015 - « Facial Expression Recognition and Analysis Challenge 2015 » – sspnet.eu/fera2015 [HA14] Md. A. Hasnat, O. Alata and A. Trémeau, « Unsupervised RGB-D image segmentation using joint clustering and region merging », Oral presentation (7.7%) in the British Machine Vision Conference (BMVC), Nottingham, UK, September 2014. [WG13] Walter S, Gruss S, Ehleiter H, Tan JW, Traue HC, Crawcour S, Werner P, Al-Hamadi A, Andrade AO, Moreira da Silva G. "The BioVid Heat Pain Database - Data for the Advancement and Systematic Validation of an Automated Pain Recognition System" In Proceedings of the IEEE International Conference on Cybernetics Lausanne, Switzerland, 2013.