2018-01083 - recognizing objects and action in the object manipulation setup

Type de contrat : CDD de la fonction publique

Niveau de diplôme exigé : Thèse ou équivalent

Fonction : Ingénieur scientifique contractuel

Niveau d'expérience souhaité : De 5 à 12 ans

Mission confiée

The goal of the project is to construct new models and learning techniques to develop and learn new powerful representations for image and video understanding. Such representation should support and advance the tasks of visual object and action recognition. Specifically, the project will focus on recognizing objects and action in the object manipulation setup. Initially the manipulation will be performed and learned in the physical simulations settings and pre-recorded real videos of object manipulation by people. Later, depending on the availability of hardware, the representations should be transferred to real robot setup.

Principales activités

The project will build on and will advance existing techniques of reinforcement learning, imitation learning and convolutional neural networks. The project may involve collaboration with PhD students and interns at the WILLOW team and should lead to a scientific publication.

Compétences

The candidate should have a PhD degree or higher in engineering or computer science, should have at least one year postdoc experience, should have a number of publications in the leading conferences and journals of computer vision and should be fluent in spoken and written English. The candidate should possess good mathematical background and programming skills.

Avantages sociaux

  • Restauration subventionnée
  • Transports publics remboursés partiellement