Multi-city street-sidewalk imagery from pedestrian mobile cameras

Abstract

This paper presents TerraFirma, a multi-city dataset which captures street and sidewalk imagery from the pedestrians' perspective. Motivated by challenges in the realm of pedestrian safety, we present a diverse and extensive dataset that provides a foundation for the design and validation of pedestrian safety systems that rely on street-sidewalk imagery. The data was collected by 9 volunteers in 4 metropolitan cities across the world. Volunteers carried mobile cameras or smartphones in a texting position, such that the rear camera was directed to the ground in front of them. TerraFirma classifies images by the material used for street/sidewalk construction in each city. The detailed description of dataset accrual is accompanied by the public release of the dataset.