A joint PhD position between 3D SAM (IMT Lille Douai / CRIStAL (UMR CNRS), France, and MICC, University of Florence, Italy)
We are looking for motivated, talented candidates for a PhD position in the area of analysis/prediction of human behaviour from temporal sequences in the wild.
This PhD program will take place at the Media Integration and Communication Center (MICC) at University of Florence (UNIFI), Italy, and the 3D SAM CRSItAL (UMR CNRS) at IMT Lille Douai, France. MICC is a pioneer laboratory in Computer Vision, and 3D SAM has a long experience in the analysis of 3D human behaviour understanding. The PhD student will work under the supervision of Prof. Stefano Berretti (MICC/UNIFI), Prof. A. Del Bimbo (MICC/UNIFI), Prof. Pietro Pala (MICC/UNIFI) and Prof. Mohamed Daoudi (IMT Lille Douai, CRIStAL).
In the last few years, we have assisted to a proliferation of methods for automatically analysing human action/behaviour by using data acquired with RGB-D cameras. Though these considerable efforts resulted in remarkable advancements in the field, existing solutions are still far from being deployable in real application contexts. There are some main reasons for this: (i) current RGB-D devices used to acquire benchmark datasets are constrained to work indoor at short distance; (ii) datasets include short sequences, where subjects show posed actions; (iii) the analysis is mostly oriented to action detection and recognition, while important tasks as interaction and prediction are not considered. In this proposal, we devise to move a step forward in the analysis of human action/behaviour, with the ultimate goal of making this field entering a more mature dimension, where concrete impact in real world applications is expected. On the one hand, we will propose innovative methods for the analysis and prediction of temporal sequences. To this end, we aim to combine geometric methods that model the temporal dimension through non-linear manifolds, with the discriminative power of Deep Learning methods. On the other, we will apply these methods to RGB-D data acquired with new depth sensors that can work outdoor and at distance. This will allow long sequences, large variability and spontaneous human behaviour. All these aspects draw a challenging and completely new scenario for RGB-D solutions that goes well beyond what proposed in the existing literature.
Strong preference will be given to candidates with experience in Computer Vision and Deep Learning, and a good knowledge of written and spoken English. The following expertise is especially considered:
The position is for duration of three years.
Candidates are invited to send their CV before June 15th2018, detailing their academic background with courses and grades during the last two years to firstname.lastname@example.org, email@example.com
(c) GdR 720 ISIS - CNRS - 2011-2018.