Integrating A Deep Learning-based Plane Detector in Mobile AR Systems for Improvement of Plane Detection

With the increasing interest in Augmented Reality (AR) technology and the enhancement of mobile device capability, mobile AR systems become very popular. In mobile AR systems, plane detection, which plays a major role in determining the location of virtual objects, often shows insufficient performance. We address this limitation in mobile AR systems by adopting a Machine Learning (ML)-based plane detector. We specifically develop a hybrid plane detection pipeline in which an ML-based plane detector and the point cloud information obtained by a mobile device are fused, showing its robust performance in plane detection for various environment conditions.