We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

New ML Datasets in Daylight Release

Formal Metadata

Title
New ML Datasets in Daylight Release
Title of Series
Number of Parts
41
Author
License
CC Attribution 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Identifiers
Publisher
Release Date
Language

Content Metadata

Subject Area
Genre
Abstract
Mapping from satellite imagery in highly populated areas can be very challenging when dense buildings and trees occlude the surface. Device location signals can provide evidence about the existence of roads in these scenarios. Grab, the leading superapp for deliveries, mobility and financial services in Southeast Asia, collects and processes GPS signals from their network of drivers. These signals are anonymized and averaged to produce vector data after some machine learning inference and centerline extraction. Based on our collaboration with Grab, we present a framework on conflating this data against OSM roads. As a result we identified 240,000 kms of new roads in 62 cities where image-based ML models have difficulty due to occlusion. The outputs are being made public through the Daylight Release. This talk was presented at State of the Map US 2022. To learn more about State of the Map US 2022, visit https://2022.stateofthemap.us/ Learn more about OpenStreetMap US at https://www.openstreetmap.us/