A Review and Case Study of Neural Network Techniques for Automated Generation of High Level-of-Detail 3D City Models

The growing interest in creating digital twins of cities has sparked a surge in the development of detailed 3D models. In this paper we examine the current state-of-the-art in generating high-resolution 3D models of cities using neural network techniques. Additionally, we showcase the outcomes of two case studies that demonstrate the practical applications of these techniques in 3D city model generation. The first case study focuses on rooftop segmentation using publicly available Swedish cadastral data, while the second case study explores façade feature extraction using Google Street View data.

Duke Scholars

Author Nikos Pitsianis Computer Science