Abstract

This bachelor thesis deals with the stereoscopic real-time representation of people using Azure Kinect and HoloLens 2. A technical challenge lies in the representation of the captured real person as a three-dimensional virtual object. Most 3D engines require the description of the person as a three-dimensional mesh. With conventional RGB-D cameras unable to create meshes, standard mechanisms of 3D engines cannot be used without further processing. A virtual 3D environment is created with the game engine Unity, which contains a 3D avatar and a virtual screen. This will then transition into a 3D environment running on the HoloLens 2. This thesis examines four technical variants to solve the problem of real time rendering: mesh generation, displacement shader, point cloud and volumetric shader. The evaluation is based on graphical quality, performance, latency and quality of the 3D model. Based on early texts we will select suitable implemented variants and conduct an additional in depth testing phase. It is shown which technical solution(s) is (are) best suited for the final application.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]