The Informatics Institute at the University of Amsterdam invites applications for a PhD position for four years, on the topic of 3D Computer Vision. The candidate will be supervised by Thomas Mensink and Arnold Smeulders.
The ultimate goal of this position is to enable 3D reasoning based on a single 2D photo. We aim to estimate the rough 3D geometry by separating the layout of objects in the scene from the global scene layout. While objects have an almost infinite number of possible configurations, the global scene layout is relatively more stable and can be cast in about 20 scene geometry types. The first research question is to define these different types and infer them from a single image alone using deep learning. Next, we focus on the local ordering of objects, to infer out-of-context objects and to describe an image based on this 3D ordering.
The research position is part of a collaboration between SRI Stanford (USA), IDIAP (Martigny, Swiss) and the University of Amsterdam to automatically infer inconsistencies among the different modalities of a video. To this end the 3D geometry delivers an important cue to match the visual and audio channel. Within the collaboration the University of Amsterdam focusses on the visual scene analysis.
Informal inquiries may be obtained from: Thomas Mensink (thomas-dot-mensink-at-uva.nl)
Additional information and application procedure:
Employer: University of Amsterdam
Expiration date: Wednesday, March 15, 2017
More information: http://www.uva.nl/en/about-the-uva/working-at-the-uva/vacancies/item/17-046-phd-candidate-in-3d-computer-vision.html