Institut National de Recherche en Informatique et en Automatique (INRIA)

The INRIA Saclay –Île-de-France research centre comprises 360 staff and 29 research teams that are all in partnerships (CNRS, Université Paris-Sud, École polytechnique, École normale supérieure de Cachan, École centrale de Paris).

Research and innovation is also booming thanks to privileged relations with the CEA (Neurospin), in the joint INRIA/Microsoft Research centre and in the Systematic Paris-Région competitiveness cluster in which INRIA research teams are heavily involved. INRIA Saclay is aiming to become a key player in establishing the Saclay plateau as a global scientific cluster.

Ecole Centrale Paris is among the most selective engineering schools in France  (Grandes-Ecoles). It was founded in 1829, was the first major engineering school to train engineers in the early days of industry and today is consistently ranked among the top three French schools of engineering.

GALEN is a joint research team between Ecole Centrale Paris and INRIA-Saclay Ile-de France, hosted at INRIA. GALEN consists of approximately 20 members, including five faculty

Role in I-SUPPORT:

INRIA has a major contribution in WP4.

  • INRIA will customize existing machine learning and localization algorithms to exploit the particularities of the ISUPPORT problem, including (i) the availability of exclusively Depth data, (ii) the multi-camera setup and (iii) the particularities of human motion in the considered constrained environment.
  • On the machine learning side INRIA intends to develop novel techniques around the problem of accurate estimation of structured and continuous data (such as the collection of positions and angles of the user‘s limbs), collectively known as ‗structured prediction’ techniques.
  • INRIA will also estimate the 3D pose of the articulated robotic manipulator using depth cues from multiple cameras based on a combination of 3D vision and machine learning techniques for object detection. INRIA will model the 3D pose of the robotic arm in terms of a chain-structured graphical model, where graph nodes encode the pose of individual arm parts, and graph edges their geometric dependencies.