Advances in Geospatial Data Enrichment and Reconstruction

The field of geospatial data analysis is moving towards increased automation and accuracy in reconstructing and enriching building and infrastructure models. Recent developments have focused on leveraging multimodal data sources, such as street-level imagery and crowdsourced data, to improve the extraction of objective building attributes and semantic descriptors. Novel frameworks and methods have been proposed to integrate diverse open datasets, automate the detection of building facades, and infer comprehensive building attributes. Additionally, there has been a push towards generating simulation-ready 3D thermal models and reconstructing view-centric CAD models from single RGB-D scans. Noteworthy papers include OpenFACADES, which introduces an open framework for enriching building profiles with objective attributes and semantic descriptors, and Thermoxels, which proposes a voxel-based method for generating simulation-ready 3D thermal models. CADCrafter is also notable for its ability to generate parametric CAD models from unconstrained real-world images.

Sources

OpenFACADES: An Open Framework for Architectural Caption and Attribute Data Enrichment via Street View Imagery

An Efficient Second-Order Adaptive Procedure for Inserting CAD Geometries into Hexahedral Meshes using Volume Fractions

An Algebraic Geometry Approach to Viewing Graph Solvability

View2CAD: Reconstructing View-Centric CAD Models from Single RGB-D Scans

Thermoxels: a voxel-based method to generate simulation-ready 3D thermal models

CADCrafter: Generating Computer-Aided Design Models from Unconstrained Images

Texture2LoD3: Enabling LoD3 Building Reconstruction With Panoramic Images

Zero-Shot Image-Based Large Language Model Approach to Road Pavement Monitoring

Built with on top of