Description:	This file contains maps of street-views from Google Maps along with a colour-coded tactile map equivalent. The tactile maps contain 6 features (streets, highways, buildings, parks, bodies of water, and medical facilities) and span 4 English-speaking countries (Australia, Canada, the UK, and the US). This dataset is intended to provide a large-scale dataset of tactile maps, and has been originally used for the training of machine learning models for tactile map generation. The main motivation was to contribute towards research and applications aimed at improving accessibility for blind and visually impaired people. The dataset is split into 4 different zoom levels (namely, levels 15, 16, 17, and 18 on Google's scale) and are all based on the same locations. At each zoom level, there is a training and a test set. Each training set consists of 5000 source-tactile image pairs, and each test set contains 500 source-tactile image pairs. In addition, a test set containing maps from 1000 cities from around the world is also included. This latter dataset can also be found for all zoom levels (15-18). All maps, both source and tactile, were retrieved using the Google Maps Static API. All maps are also copyrighted by Google. For all use-cases, including reusing, modifying, or extending this dataset or any of the images, please ensure to follow Google's Terms of Service as well as the Terms of Service for Google Maps.
Authors:	Hobson, David G.; Carleton University; 0009-0000-7911-1501 Komeili, Majid; Carleton University; 0000-0002-4695-3072
Keywords:	Google Maps Accessibility Maps Street-views Tactile maps Machine learning Artificial intelligence Blind and visually impaired Maps for the blind
Field of Research:	Computer and information sciences > Artificial intelligence (AI) > Computer vision in artificial intelligence Computer and information sciences > Artificial intelligence (AI) > Machine learning
Publication Date:	2023-09-21
Publisher:	Federated Research Data Repository / dépôt fédéré de données de recherche
Funder:	Natural Sciences and Engineering Research Council of Canada
URI:	https://doi.org/10.20383/103.0797