Postdoc Researcher - Linking Perception and Manipulation of Non-Rigid Objects through Deep Learning

Application deadline:

The goal of this project is to develop novel deep learning techniques to capture the geometry of dressed persons from a single RGB image. Current approaches to reconstruct human shape consider low rank parametric models that can only reconstruct the shape of the naked body of very tight clothes. We will investigate different representations of the 3D geometry to train Deep Networks and capture both the geometry of the clothes and the underlying shape of the body. The candidate will be supervised by Dr. Francesc Moreno and Prof. Carme Torras , and she/he will work in cooperation with the members of the Perception and Manipulation group at IRI. The candidate is expected to have good prior knowledge in deep learning libraries, preferably pytorch. A one-year postdoc position under the MdM Strategic Research Programme will be fully covered starting as soon as possible. The salary will be around 36000 € gross per year. In addition, 5500 € per year will be available for research expenses of the postdoc (travels, registrations, equipment, training, etc).