Building Floor Number Estimation from Crowdsourced Street-Level Images: Munich Dataset and Baseline Method

TL;DR

This work addresses the challenge of obtaining building floor numbers by leveraging unrestricted street-level imagery. It introduces an end-to-end deep learning framework with a classification-regression baseline and a data-generation pipeline that combines Mapillary and self-captured images for Munich, paired with CityGML LoD2 ground truth. The proposed HTTC/HYB approach achieves state-of-the-art performance among baselines, with $81.19\%$ exact accuracy and $97.90\%$ within $±1$ floor, demonstrating scalable potential for enriching 3D city models with vertical attributes. The Munich Building Floor Dataset and baseline network pave the way for broader urban informatics applications, while highlighting remaining challenges in data scalability and cross-city generalization.

Abstract

Accurate information on the number of building floors, or above-ground storeys, is essential for household estimation, utility provision, risk assessment, evacuation planning, and energy modeling. Yet large-scale floor-count data are rarely available in cadastral and 3D city databases. This study proposes an end-to-end deep learning framework that infers floor numbers directly from unrestricted, crowdsourced street-level imagery, avoiding hand-crafted features and generalizing across diverse facade styles. To enable benchmarking, we release the Munich Building Floor Dataset, a public set of over 6800 geo-tagged images collected from Mapillary and targeted field photography, each paired with a verified storey label. On this dataset, the proposed classification-regression network attains 81.2% exact accuracy and predicts 97.9% of buildings within +/-1 floor. The method and dataset together offer a scalable route to enrich 3D city models with vertical information and lay a foundation for future work in urban informatics, remote sensing, and geographic information science. Source code and data will be released under an open license at https://github.com/ya0-sun/Munich-SVI-Floor-Benchmark.

Figures (15)

• Figure 1: Examples of SLI used in this study. (a)-(d) Images from Mapillary, showcasing diverse building facades captured from crowdsourced imagery, requiring cropping to isolate buildings and filtering to remove low-quality building images. (e)-(j) Self-captured images, focusing on underrepresented high-rise buildings, with centered framing for clear visualization of facades.
• Figure 2: (a) Coverage of LoD2 building models in Munich, and (b) Example of LoD2 building models in Munich.
• Figure 3: Distribution of Floor Counts in Munich LoD2 CityGML Database (Log Scale), with building counts labeled on each bar.
• Figure 4: Scatter plots of building floor and height for selected building functions and roof types.
• Figure 5: Examples of building detection (a) and cropping (b).