Generating spatially-coherent 3D representations of real estate properties from multi-source consumer-grade data

Abstract

The <b>Real Estate in 3D</b> project investigates how spatially coherent 3D representations of rental properties can be generated using consumer‑grade data commonly found in real estate listings. The project addresses the limitations of current property search processes, where decisions are often based on fragmented visual and spatial information. In collaboration with the Swiss start‑up <b>SwissPlatz sàrl</b>, the research explores deep‑learning‑based reconstruction, multimodal data fusion, and computer vision techniques to transform photos into unified 3D models. The approach avoids costly professional scanning solutions and requires no additional data acquisition. The results form the foundation for a scalable, affordable 3D property exploration service that improves spatial understanding for prospective tenants.