AI Insight
This study presents a deep learning approach using vision transformers (DINOv2 architecture) to create accurate building maps for smart cities by combining imperfect open-source maps with radio frequency data from wireless networks. The method achieved 65.3% macro IoU on synthetic data and 64.9% on real-world Oslo data, significantly outperforming baseline methods that relied on either RF data alone (37.3%), erroneous maps alone (40.1%), or non-AI fusion approaches (42.2%). The approach demonstrates that AI-based fusion of multiple imperfect data sources can substantially improve urban mapping accuracy compared to using individual data modalities.
Why it matters
This technology could enable more accurate and cost-effective mapping of urban environments for smart city applications by leveraging existing wireless infrastructure data rather than relying solely on expensive aerial imaging or manual surveys. The ability to improve mapping accuracy even when using imperfect or noisy data sources makes this approach particularly practical for real-world deployment in developing urban digital twins and infrastructure planning.
Understand the Science
arXiv:2508.03736v2 Announce Type: replace-cross
Abstract: In this paper, we present a deep learning-based approach that integrates the DINOv2 architecture to improve building mapping by combining (possibly erroneous) maps from open-source platforms with pervasive radio frequency (RF) data collected from multiple wireless user equipments and base stations. Unlike prior methods, our approach leverages a vision transformer-based architecture to jointly process both RF and map modalities within a unified framework, effectively capturing spatial dependencies and structural priors for enhanced mapping accuracy. For the evaluation purposes, we employ a synthetic dataset co-produced by Huawei. To address the challenges associated with real-world data imperfections, we introduce controlled noise to its RF data so as to simulate real-world conditions. Additionally, we develop and train a model that leverages only aggregated path loss information to tackle the mapping problem. We measure the results according to three performance metrics: the Jaccard index (intersection over union, IoU), the Hausdorff distance, and the Chamfer distance. Our design achieves a macro IoU of 65.3%, significantly surpassing (i) the erroneous maps baseline, which yields 40.1%, (ii) an RF-only method from the literature, which yields 37.3%, and (iii) a non-AI fusion baseline that we designed which yields 42.2%. The comparative evaluation highlights the limitations of relying solely on RF data or on spatial data, as well as the effectiveness that AI can have on fusing data towards enhancing smart city mapping accuracy. We further validate our method on real-world data from the Oslo region, complementing the synthetic evaluation with a real deployment setting, where our best fusion model reaches 64.9% macro IoU. We additionally outline a strategy for deploying the model over larger areas by tiling the region with overlapping windows.