3D PERCEPTION FOR AUGMENTED REALITY

Abstract

Augmented Reality (AR) is a technology that integrates 3D virtual objects into a 3D real environment in real-time, which has three key features: 1) combining virtual and real objects, 2) registering real and virtual objects with each other in 3D, 3) interacting in real time. AR is a novel way for digital content presentation that blurs the boundary between virtual and realistic worlds and provides a more immersive experience. It has been applied in many fields, including medicine, education, architecture, marketing, maintenance and assembly process. However, the experience of AR is still far from ideal. In this thesis, we mainly focus on improving 3D perception algorithms for AR, especially for mobile AR, including efficient depth estimation and 3D human pose estimation algorithms. 3D perception refers to perceiving 3D information of the surrounding environment and objects.