Real Estate in 3D, FHNW School of Computer Science

Finding a new home is one of life’s major decisions, yet it is often made based on fragmented information. Prospective tenants typically rely on short descriptions, a small number of photos taken from arbitrary angles, and a brief, crowded viewing. Important spatial details are hard to compare, dimensions are difficult to estimate, and impressions fade quickly. As a result, decisions with long-term consequences are frequently made without a reliable understanding of the space.

While professional solutions for capturing buildings in 3D exist, such as LiDAR scanning and dedicated photogrammetry setups, they remain expensive and impractical for the everyday rental market. This creates a clear gap between what is technically possible and what is realistically available to tenants.

This project aims to bridge that gap. In collaboration with SwissPlatz, a Swiss start up focused on automating and improving the relocation experience, the Institute of Interactive Technologies (FHNW) investigates how spatially coherent 3D representations of real estate properties can be generated using only consumer grade data, such as photos and floor plans that already exist in rental listings.

The project explores a multi stage processing pipeline that combines deep-learning-based 3D reconstruction, multimodal data fusion, and computer vision techniques. Scattered, heterogeneous inputs are transformed into unified spatial models that can be explored intuitively, without requiring additional hardware, professional capture, or new data collection.

The outcomes provide SwissPlatz with the scientific and technical foundation for a scalable and affordable 3D property exploration service, offering prospective tenants a spatial understanding that comes much closer to actually being there.