Traffic accidents are a major global concern, highlighting the need for advanced traffic analysis and predictive techniques. The emergence of crowdsourced street view imagery (SVI) platforms has transformed public participation in collecting urban data, initiating the development of human-centered traffic analytics. This article investigates the relationship between visual urban understanding derived from Mapillary SVI and the frequency of urban traffic accidents. We analyzed SVI’s visual complexities using the Mask2Former image segmentation model and provided spatial reasoning of the urban objects (e.g., cars, buildings, trees) by estimating visual distances between those objects and the drivers using Dist-YOLOv5. These distances were encapsulated as edge weights, with urban objects and the drivers as nodes, creating human-centered graphs at each SVI location. We propose a framework that integrates a graph-based deep learning approach, GAT-LSTM, to capture the spatial-temporal dynamics of these urban objects for modeling traffic-accident frequency. Our results indicate that this model outperforms a traditional machine learning method by over 70 percent in mean absolute percentage error (MAPE) and demonstrates superior performance compared to other deep learning-based methods. Additionally, we introduce a two-step Explainable AI (XAI) method to identify key factors associated with roads with higher traffic accident rates, thereby improving the interpretability and practicality of our research for understanding the urban safety environment.

City landscapes viewed through home windows influence quality of life, yet perceptions of actual window views at the urban scale remain understudied. This study presents an approach for large-scale mapping of perceptions using 12,334 window view images (WVIs) collected from actual residential properties listed on real estate platforms in Wuhan, China, representing a rarely explored form of urban view imagery that offers advantages over the rendered or simulated window views commonly examined in previous studies. Through a non-immersive virtual reality platform, we collected 27,477 pairwise comparisons across six perceptual dimensions (e.g. preference) from 304 participants based on 499 WVIs. A hybrid neural network model was trained to predict human perceptions of all crowdsourced WVIs and map their spatial distribution. Results reveal significant spatial autocorrelation with distinct hot and cold spots across the whole city. Floor level strongly influences human perceptions: while higher floors offer more preferred and extensive window views, lower-floor windows provide residents with quiet and vivid views. An inference model further shows that window view composition matters considerably: high ratios of sky, trees, and low-rise buildings enhance people’s preferences and perceptions of vividness, whereas high ratios of high-rise buildings increase perceptions of monotony and oppression. Importantly, these effects are non-linear: the excessive presence of certain elements can alter their impact on human perception. This work advances urban-scale understanding of residents’ visual experiences and offers a transferable, human-centric method to inform urban planning and design aimed at improving the visual quality of window views.

Globally consistent semantic digital twins require centimeter-accurate and geographically transferable 3D facade segmentation. However, progress in facade parsing is limited by the lack of large-scale, standardized benchmarks for evaluating cross-domain generalization. Existing datasets are geographically narrow, semantically inconsistent, or insufficiently precise. We introduce UnderOneFacade, the largest cross-country and cross-continent 3D facade benchmark to date, comprising centimeter-accurate point clouds with hierarchical, harmonized, and architecturally grounded semantic labels totaling 2.7 billion annotated points. Through a systematic evaluation of representative point-, graph- and transformer-based architectures, we show that current methods struggle to recognize fine-grained architectural elements and degrade significantly across geographic domains, with the best models achieving only up to 33 IoU on the fine-grained LoFG3 benchmark. By combining geometric precision with standardized semantics at unprecedented scale, UnderOneFacade establishes a rigorous benchmark for developing robust and transferable 3D segmentation models. The dataset, evaluation scripts, and pretrained models will be released upon publication.

Where we live profoundly shapes our health, with urban environments playing a critical role in shaping population health outcomes. As health disparities persist within and between cities, ensuring equitable urban design has become critical to advancing population well-being. Yet most studies focus on case studies of single cities and overlook differences between general, physical and mental health dimensions, limiting our understanding of how urban factors shape health outcomes at scale. Here, to address this gap, we integrate census-tract-level health data, crowdsourced geospatial information and deep learning to identify urban features associated with general, physical and mental health across the most populous urban areas in the United States. Our analysis reveals distinct associative relationships through which urban contextual and socioeconomic factors shape health outcomes. We identify the ranked importance of urban determinants for each health dimension, along with cross-cutting factors that consistently matter. Our findings suggest that urban service enhancements in low-income neighbourhoods are associated with 100–462% greater health gains over high-income areas. Furthermore, we find strong links between the heterogeneity of urban spatial patterns and both health and income inequalities. Overall, our findings highlight strong associations between equitable access to urban services and coherent city planning with observed patterns of population health inequalities across cities.