We are all familiar with Google Maps, the mapping solution run by the search giant, but did you know there are other mapping solutions like Bing Maps, HERE WeGo, and OpenStreetMap? OpenStreetMap (OSM) is unique among these because it is open source and doesn"t have the backing of a big tech company with lots of resources.
One of the big issues with the lack of resources is that it means OSM relies heavily on manual updates to its maps whereas companies like Google already use technology to do this automatically. Now, however, researchers have created an end-to-end machine-learning solution called DeepMapper which lets you select areas on the latest satellite imagery and convert that data in those images to update OSM maps.
In an accompanying research paper, the developers of DeepMapper compared their tool to others developed by Microsoft, Mapbox, Facebook, IBM, and OSM itself. It looked to see whether they had road, building, and other object detection capabilities, geo-referencing, and the ability to automatically update OSM.
In some tools, object detection was good but geo-referencing and updating OSM were lacking, in other cases object detection was what was lacking. With DeepMapper, it can detect roads and buildings, but not other objects, and it supports geo-referencing and can update OSM - this makes it an ideal tool for updating OSM based on the latest satellite imagery. DeepMapper also follows OSM"s Automated Edits Code of Conduct ensuring that its generated updates contribute to the integrity and reliability of the platform"s database.
As tech giants use more and more sophisticated technologies to improve their mapping software, hopefully, projects like OpenStreetMap can keep pace with the help of tools like DeepMapper. There will still be a place for people to contribute fixes to maps as is common now but these tools will contribute the vast majority of changes.
If you"re interested in taking a look at the open-source code, it is available on GitHub.