TerraFirma Dataset

TerraFirma is a first-of-its-kind database that captures street/sidewalk imagery from a pedestrian's perspective. It consists data from four cities: New York, London, Paris, and Pittsburgh. The data was collected in the wild, on real smartphones, on people’s daily walking paths. For each city, the data is classified in 2 classes: sidewalk and street. Each class has labeled images for materials constituting that class. The images are labeled manually based on the material in the center of the frame. Labels are aggregated; for example: sand-colored, pigmented, and unpigmented concrete is labeled as concrete.


City Sidewalk material Street material
New York Concrete; Tiles; Tinted Asphalt
London Tiles; Brick Asphalt; Brick
Paris Asphalt Asphalt; Brick
Pittsburgh Concrete; Brick Asphalt; Brick

Download Now

Reference

Recognizing Textures with Mobile Cameras for Pedestrian Safety Applications [PDF]

Shubham Jain and Marco Gruteser
IEEE Transactions on Mobile Computing

BibTeX

@ARTICLE{8454285,
author={S. {Jain} and M. {Gruteser}},
journal={IEEE Transactions on Mobile Computing},
title={Recognizing Textures with Mobile Cameras for Pedestrian Safety Applications},
year={2018},
doi={10.1109/TMC.2018.2868659},
ISSN={1536-1233},}