Our research is focused on 3D perception for
artificial systems. The goal is to develop the theoretical and
practical fundamentals that will allow us to build seeing systems
for dynamic and unconstrained environments. To this aim, we
make use of the most recent sensor technology, such as 3D cameras,
laser range finders or omnidirectional cameras and we draw inspiration
from a wide range of scientific communities, including Computer
Vision, Robotics, Machine Learning, Computer Graphics and Physics.
In one of our current projects, we study the marker-less human motion capture problem, that is, the problem of estimating the 3D motion of moving human subjects in real-time from a stream of camera images. If such motion capture technology were to become convenient, cheap, and applicable in natural environments, a whole range of applications would become possible, such as intuitive human-machine interaction, smart surveillance, character animation, virtual reality or motion analysis.
Human motion capture, on the other hand, is an exceptionally hard perception problem due to its high dimensionality, inherent uncertainty and ambiguity, which makes it an excellent research target.
For more information on our approaches and other projects, please don't hesitate to contact us.