An AI-based system offering automatic DR-enhanced AR for indoor scenes

Abstract

In this work, we present an AI-based Augmented Reality (AR) sys- tem for indoor planning and refurbishing applications. AR can be an important medium for such applications, as it facilitates more effective concept conveyance, and additionally acts as an efficient and immediate designer-to-client communication channel. However, since AR only overlays, and cannot replace or remove, our system relies on Diminished Reality (DR) to support deployment to real-world already furnished indoor scenes. Further, and contrary to the traditional mobile AR application approach, our system offers on-demand Virtual Reality (VR) viewing, relying on spherical (360o) panoramas, capitalizing on their user-friendliness for indoor scene capturing. Given that our system is an integration of different AI services, we analyze its performance differentials concerning the components comprising it. This analysis is both quantitative and qualitative, with the latter realized through user surveys, and provides a complete systemic assessment of an attempt for a user-facing, automatic AR/DR system.