Robocodes: Towards Generative Street Addresses from Satellite Imagery

CVPR 2017

Abstract

We describe our automatic generative algorithm to create street addresses (Robocodes) from satellite images by learning and labeling regions, roads, and blocks. 75% of the world lacks street addresses [12 ]. According to the United Nations, this means 4 billion people are ‘invisible’. Recent initiatives tend to name unknown areas by geocoding, which uses latitude and longitude information. Nevertheless settlements abut roads and such addressing schemes are not coherent with the road topology. Instead, our algorithm starts with extracting roads and junctions from satellite imagery utilizing deep learning. Then, it uniquely labels the regions, roads, and houses using some graph- and proximity-based algorithms. We present our results on both cities in mapped areas and in developing countries. We also compare productivity based on current ad-hoc and new complete addresses. We conclude with contrasting our generative addresses to current industrial and open solutions.

Latest Publications

Sustainable AI: Environmental Implications, Challenges and Opportunities

Carole-Jean Wu, Ramya Raghavendra, Udit Gupta, Bilge Acun, Newsha Ardalani, Kiwan Maeng, Gloria Chang, Fiona Aga Behram, James Huang, Charles Bai, Michael Gschwind, Anurag Gupta, Myle Ott, Anastasia Melnikov, Salvatore Candido, David Brooks, Geeta Chauhan, Benjamin Lee, Hsien-Hsin S. Lee, Bugra Akyildiz, Max Balandat, Joe Spisak, Ravi Jain, Mike Rabbat, Kim Hazelwood

MLSys - 2022