10 Million Street Views

Street-level imagery such as Google Maps Street View panoramas has become a pivotal resource for many researchers as it can provide a unique perspective on built environments. The ability to access and analyse comprehensive street-level imagery provides researchers with a powerful tool for exploring and understanding urban environments. 

Accessing comprehensive street level imagery at scale can be difficult, expensive and time consuming. Which is why the Urban Analytics Lab at the National University of Singapore (NUS) has introduced the Global Streetscapes project. The NUS Global Streetscapes project provides extensive coverage of urban street level imagery, with 10 million street-level images across 688 cities worldwide, enriched with over 300 attributes. 

The dataset includes images from both Mapillary and KartaView, two crowdsourced 'street view' platforms that offer a diverse range of street-level imagery. Each image in the dataset is annotated with attributes, such as the type of road, weather conditions, and the mode of transportation used to capture the image. These annotations enable researchers to filter and select the images that are the most relevant to their specific studies, for example for evaluating walkability or mapping the levels of urban greenery.

The Global Streetscapes project has even pre-computed some of the evaluations of street view imagery that researchers commonly use, such as the green view index, which ranks the ratio of vegetation pixels in an image to the total number of pixels. 

The NUS Global Streetscapes project is free to use. The project provides open access to a comprehensive dataset of 10 million street-level images enriched with extensive metadata, as well as the code and documentation necessary for using and extending the dataset. All the code and documentation for the project can be found on the Global Streetscape GitHub page and the dataset itself can be accessed on the project's Hugging Face page.

Hat-tip: Map Channels