Unsupervised Scene Understanding by exploring Depth and Motion Cues

Christoph Reich (Ph.D. Student)

Scene understanding is a common computer vision task involving object recognition and semantic reasoning. Recently, much progress has been made in supervised scene understanding. However, the absence of labeled data in novel and complex scenarios poses particular challenges. The goal of this project is to tackle scene understanding with largely reduced supervision. In particular, we aim to integrate knowledge about motion as well as depth as inductive biases for unsupervised scene understanding.

Primary Host:	Daniel Cremers (Technical University of Munich)
Exchange Host:	Christian Rupprecht (University of Oxford)
PhD Duration:	01 December 2023 - 01 December 2027
Exchange Duration:	01 February 2025 - 31 July 2025 - Ongoing