Applications of Street View Images based on Artificial Intelligence: A Comprehensive Survey

Abstract

Street View Images (SVI) are high-resolution, geo-referenced panoramas that capture real-world environments. Integration of Artificial Intelligence (AI) with SVI enables automated analysis for a range of urban applications including object detection, semantic segmentation, text recognition, scene understanding, and socioeconomic prediction. More than 25 recent AI based SVI studies covering applications in crime prediction, building attribute classification, sidewalk inventory, land price estimation, and environmental monitoring were screened for the survey. Across domains, deep learning architectures such as ResNet, ConvNeXt, and Vision Transformers consistently outperformed traditional machine learning models, with reported accuracies up to 94% for classification and R2 values between 0.62–0.83 for prediction tasks. SVI augmented with other data sources like satellite imagery and Global Information System (GIS), enhanced the model performance and contextual understanding. The Findings highlight the dominance of convolutional and transformer-based networks, emerging interest in graph neural networks, and the need for generalized models and diverse datasets to advance SVI research.