The axis of facial expression recognition (affective computing) consists of facial expression recognition in the presence of occlusions, facial cue detection by modeling local and global movements [PhD – R. Belmonte], motion descriptors for modeling macro and micro facial expressions from video in a real environment (in-the-wild) [PhD – B. Allaert], dynamic face recognition in audiovisual programs [PhD – R. Auguste], and facial recognition by multilayer perceptrons.

The axis includes the recognition of human actions in a personal environment [PhD – Y. Benabbas] and crowds  [PhD – M. H. Sharif] (e.g. detection of abnormal events) and the detection and tracking of gaze. On this last point, taking into account the symmetry constraint of the face for pose detection based on global appearance [PhD – A. Dahmane], cylindrical or symmetrical approach of the face [PhD – A. Lablack], bimodal face recognition by fusion of visual and depth features [PhD - A. Aissaoui], posture estimation by learning based on Gabor descriptors and independent of facial cues in a controlled environment. Gaze estimation is a geometric projection, developed to extract the region of interest of a person in a targeted scene. Finally, the extracted data is analyzed to derive gaze points, visual map and visual pathways.

The axis of automatic indexing by content includes representation by visual words and association of regions for learning in image databases [PhD – I. Elsayad], clustering by projection on random axes, optimization of multidimensional indexing [PhD – T. Urruty], Fourier descriptors for modelling shapes for indexing and searching images by content, and combination of visual and textual descriptors for information retrieval [PhD – M. Bouet]

The axis develops Markovian clustering methods [S. Mongy], semantic metadata integration methods based on linguistic, semantic and structural mapping of standard multimedia metadata schemas [PhD - Samir Amir], ontology-based analysis of human interactions on the web [PhD - Mehdi Adda], temporal and interactive synchronization of multimedia documents based on temporal petri nets [PhD – K. Hadouda], [PhD - A. Ghomari], extraction of business process models from event logs modeling human interactions in business processes [PhD – N. Ihaddadene], and prediction of web paths in the context of the legal deposit of the web archives of the National Institute of Audiovisual [PhD – Y. Hafri].

This is the current research axis, and for the next 5 years. The axis of bio-inspired approaches concerns impulse neural networks or deep neural networks for the detection and spatio-temporal tracking of objects or phenomena, with applications such as semantic segmentation, recognition of static and dynamic objects in an indoor environment, analysis of atmospheric phenomena, and recognition of human actions in an indoor environment via a home assistance robot, analysis of human behavior from video by studying motion orientation in videos, image processing and segmentation for site surveillance, detection of abnormal events in video, analysis of optical flows by virtual line for pedestrian counting, and prediction by time series from events in video.