We are releasing 10.4k turn restriction detections located across 5.2k intersections for the OpenStreetMap community.
Map
Raw GeoJSON data
These turn restriction detections were sourced by applying our machine learning computer vision models to Bing Streetside imagery. Mapbox has acquired the right to contribute these detections to OSM.
“Detections” are not really the same thing as turn restrictions. Rather, they are suspected, but unverified turn restrictions. We are planning to manually verify these turn restrictions, and our data team is working through the list, but the first pass produced thousands of candidates, so verification will take a while. We do think these can be useful to the OSM community, who may wish to move faster on improving OSM based on these detections.
To be clear, the Streetside imagery is not available for direct editing by the community at large. However, subject to its terms and conditions, it can be used for personal reference for examining the source of the detections, so we have included URLs to the image associated with each detection.
Examples from the Streetside Imagery API:
Discussion
نظر از CloCkWeRX در 2 فوریه 2018 ساعت 00:19
How easy would it be to fork improveosm.org and add these + the mapillary street signs as a layer in that version of ID? That could help these go from “probable” to “almost certain” where there’s multiple hints from GPS data and imagery
نظر از Glassman در 2 فوریه 2018 ساعت 00:36
Is the node in the map supposed to be selectable to see the detection results?
نظر از Baloo Uriza در 2 فوریه 2018 ساعت 04:18
I’m looking forward to this going into the Improve OSM plugin.
نظر از RobJN در 2 فوریه 2018 ساعت 20:14
“We” is Mapbox right? Anyway good stuff. Anything planned for outside of the US?
نظر از Carnildo در 3 فوریه 2018 ساعت 01:02
I don’t know how much use this will be. I checked all 15 detections in Eastern Washington, and only one of them was not already mapped through road geometry or turn restrictions.
نظر از SK53 در 4 فوریه 2018 ساعت 21:19
I only checked one in Pittsburgh, and that also appeared to be mapped.