Geometric and Semantic Modeling from RGB-D Data

Abstract

With the increasing availability of RGB-D cameras, using RGB-D data for geometric and semantic modeling has received significant interest in recent years. Geometric modeling aims to build an accurate geometric representation for 3D objects or scenes, whereas semantic modeling focuses on analyzing and understanding semantic objects in the captured scenes. They have many applications ranging from robotic navigation to VR/AR. In this chapter, we will overview recent efforts on this research topic, in particular, research using advanced machine learning techniques, exploiting the complementary characteristics of geometry and image information in RGB-D data, and incorporating prior knowledge.