Interactive Media Systems, TU Wien

Smart 3D Geometry Understanding within a Dynamic Large Triangulated Point Cloud

Master Thesis

About This Topic

The aim of this master thesis is to enable automatic object detection, starting from planar surfaces to more complex objects, within a streamed dense point cloud that is triangulated for live immersive scene exploration using virtual reality gear. To detect obstacles within the dense 3D reconstruction, computer vision / deep learning algorithms need to be applied to understand the scene geometry for subsequent highlighting and collider computation.

Hardware: HTC Vive, Cyberith Virtualizer Development Environment: Unreal Engine, Steam VR (Vive), Computer Vision / Machine Learning libraries

More information of the project: