Abstract
Cities are growing, and at a much faster pace in less developed countries compared to developed countries. Unlike developed countries, however, that typically have updated data to support important decisions about their cities, cities in less developed countries are not as fortunate. To help overcome this data poverty gap, new emerging sources of data on cities are increasingly becoming available in less developed countries. This study focuses on one such source, volunteered street view imagery (VSVI) captured from dashboard cameras and mobile devices in vehicles. Our goal is to understand the types of information that can be extracted from such data with a view towards understanding how they can be used to support strategic planning initiatives and policies in cities. Using the city of Bengaluru in India as a case study, we extracted 29,000 street view images and use a state-of-the-art image segmentation model to classify images into 36 street classes. Our results show mixed accuracies in the type of features that can be extracted from images owning in large part to the complex nature of cities. In the future, we plan to further refine the automated extraction of features from VSVI to produce a point of interest database on cities.
How to Cite:
, E., , R. & , O.,
(2022) “Understanding City Dynamics Through Street View Images: A Deep Learning Perspective”,
Journal of Student-Scientists' Research 3.
doi: https://doi.org/10.13021/jssr2021.3218
Files
Downloads are not available for this article.