Urban visual perception varies by weather conditions and functional zone types

This dataset is built from a large-scale collection of Google Street View (GSV) imagery sampled systematically across ten global cities. Using a 50-metre spatial grid, we extracted four directional images (front, back, left, right) at each sampling point within the built-up areas of London, Frankfurt, Melbourne, Singapore, New York, São Paulo, Dubai, Johannesburg, Hong Kong, and Tokyo. All images were captured from official GSV panoramas (2023–2024 updates) and screened to remove tunnels, night scenes, user-uploaded photos, and low-visibility views. The resulting dataset provides a clean, high-resolution baseline of real-world streetscapes representing diverse climates, urban forms, and socio-spatial contexts.