Description
Date depot: 11 octobre 2019
Titre: Apprentissage continu explicable pour conduite autonome
Directeur de thèse:
Raja CHATILA (ISIR (EDITE))
Encadrante :
Natalia DIAZ RODRIGUEZ (U2IS (EDX))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Intelligence artificielle
Resumé:
This thesis seeks to improve the scene understanding and the object recognition ability of an autonomous system in a complex environment using a combination of symbolic and deep learning.
The goal of this work is to develop a pipeline of Relational Deep Learning able to improve 3D scene recognition in autonomous driving contexts, thanks to a better understanding and explicability of current Artificial Intelligence methods. This visual reasoning pipeline will need to use different vision formats (photos, videos) to quickly understand autonomous driving scenes.
Doctorant.e: Bennetot Adrien