Consistent, Privacy-preserving Scenes from Multiple Depth Views and Ultra-Low-Resolution RGB images

Thesis Degree:
MasterThesis Advisors:
Xuying HuangThesis Description:
We plan to re-implement the method presented in [1], extending the network to take both depth maps and ultra-low-resolution RGB images as inputs. We will then deploy it on a real robot to reconstruct a practical, real-world 3D scene.
Thesis Start Date:
September 1, 2025Thesis Requirements:
Enrolled in computer science or similar MSc program in and around Bonn/Cologne
Familiarity with computer vision and mobile robot
Experience with diffusion and generative models is a plus
Programming experience with C++, Python
Thesis Related Work:
1. Z.Liu et al. “MOSAIC: Generating Consistent, Privacy-Preserving Scenes from Multiple Depth Views in Multi-Room Environments”, ICCV, 2025