Context:
Understanding the functions of the skeletal structures of vertebrates
in motor control and adaptive
strategies for rest posture and locomotion activity is a key issue in
several research domains.
However, biomechanical expertise reports that current techniques of
geometrical prediction of the skeleton structures from 3D external
markers present limitations in
accuracy due to the fact that flesh and muscles are not rigidly
connected to the bones structure.
The EVASION research group on computer graphics at INRIA is leading a
project on 3D characterization of the relationships between the
articulated skeleton of vertebral animals and the deformation of soft
tissues (project
ANR-Kameleon). This project focuses on the locomotion
of small quadruped
animals. It is held in cooperation with the National Museum of Natural
History in Paris, the biomechanical department of the Rennes University
(LPBEM laboratory) and the biomedical department of University
Descartes Paris (LNRS laboratory).
Work profile:
We aim to combine skeleton motion data captured with an X-ray
videography system and surface deformations measured with a structured
light 3D scanner. Standard skinning methods are known to rely on
unrealistic
approximation of the skin deformation. Using real data collected on
animals, the goal of this postdoctoral project is to evaluate the
limitations of these methods and investigate a new approach. In
particular, methods inspired by machine learning will be considered,
such as Lewis et al. in 2000 (“Pose space deformation: A unified
approach to shape interpolation and skeleton-driven deformation”,
SIGGRAPH’00), Kry et al. in 2002 (“Eigenskin: Real time large
deformation character skinning in graphics hardware”, SCA’02) or Mohr
et
Gleicher in 2003 (“Building Efficient, Accurate Character Skins from
Examples”, SIGGRAPH’03). Works by Allen et al. in 2002 have proved the
efficiency of the use of static meshes acquired from laser range
scanner (“Articulated body deformation from range scan data”,
SIGGRAPH’02). This postdoctoral project aims at developing such an
approach for the dynamic case of an animal in motion.
Requested skills: Computer Animation, with an emphasis on skinning method and free-form-deformation techniques. Knowledge of Machine Learing will be a plus.
Technical skills: C++ programming, Matlab. Knowledge of Maya software will be a plus.
Language: most exchanges in the project are in English and French. Spoken and written knowledge of English is mandatory. Knowledge of French will be a plus.
Application:
Please send a detailed curriculum vitae (maximum
4 pages), a motivation letter and at least one letter of reference from
the PhD advisor. Application should be sent by email to
lionel.reveret(at)inria.fr or by regular mail to the following address: