Generating 3D Spaces from 2D Videos using AI for Motion Capturing and Virtual Reality Engines

Abstract

This thesis explores the way to convert human movement to an asset that can be later used in virtual reality engines. Starting from collection of raw data from user, using ai to capture the movement. Exploring the ways of the data collection using different setups for better pose estimation. Continuing with making data more precise by using various techniques. Reconstructing the environment by determining static and dynamic objects. And finally getting an output of a format that is compatible with virtual reality engines.