3D pose and shape from monocular images
Place: Large Lecture Room.
Affiliation: Institut de Robòtica i Informàtica Industrial, Barcelona, Spain.
In this talk I will review my principal research topics in the past few years. I will start presenting linear formulations to efficiently estimate pose for calibrated and non-calibrated cameras. Then I will show that these formulations are also applicable to estimate the shape of deformable objects or the 3D pose of the human body from single images. Finally I will introduce our recent results on sequential estimation of non-rigid shape and camera motion from video, and new contributions on descriptor computation.