Offre en lien avec l’Action/le Réseau : – — –/Doctorants
Laboratoire/Entreprise : Greyc
Durée : 12 mois
Contact : alexis.lechervy@unicaen.fr
Date limite de publication : 2020-12-31
Contexte :
The detection of anomalies or out-of-domain data has become an important issue in machine learning, especially in the field of video surveillance. It has indeed become important to be able to detect unusual events in video sequences in order to anticipate and predict potentially dangerous situations. The work proposed in this position will be carried out within the context of an ANR project related to the preparation for the Paris 2024 Olympic and Paralympic Games. Its objective is to be able to design digital tools allowing the detection of generic unusual situations in videos from surveillance cameras in order to effectively alert an operator. This project is part of a collaboration with IDEMIA, EVITECH and CERAPS..
Sujet :
Several approaches have been proposed in the literature to perform anomaly detection. Among these approaches, a set of works is interested in the use of auto-encoders for this task. On the other hand, the use of multimodal data in modern machine learning approaches has led to improved performance on various problems such as image classification. During this post-doc, we will study the possibility of creating multimodal auto-encoders for anomaly detection. For this purpose, we plan to extend our work on multimodal fusion to the auto-encoder. The modal fusion will be performed on levels adapted to each modality. Further work will focus on the use of the proposed solution for anomaly detection in surveillance videos with, for example, applications to the ShanghaiTech, CUHK Avenue and UCSD pedestrian bases.
Profil du candidat :
PhD in Machine learning or equivalent,
Formation et compétences requises :
– Experience in Deep learning and in associated frameworks (Pytorch, Tensorflow…),
– Good experience in publishing scientific articles.
Adresse d’emploi :
6 Boulevard Maréchal Juin,
14000 Caen
France
Document attaché : 202009290821_job_detection_anomaly.pdf