I want update signage and those sorts of items in my local area from the Mapillary objects which I can verify are correct. however I’m unsure whether Mapillary automatically adds these items to OSM or not.
I’m assuming they don’t?
If I’m not mistaken it’s not allowed to add data to OSM automatically. @boris do you know this?
BR, Yaro
Mapillary does not add anything to OSM data automatically. This would qualify as an automated data import and would need to pass OSM community guidelines and review. Which it would not pass because most user-uploaded GPS is inaccurate and noisy and detections themselves have too many errors. Importing such would end up adding a lot of incorrect data to OSM. So it would at least need a manual human review. Therefore, if you do not see something mapped/added on OSM and want to use Mapillary to add it, you can go ahead and do so. No import or automation will overwrite or duplicate anything. There currently aren’t even any tools or editors to map from Mapillary quickly and efficiently (only partial solutions for preset tasks like Pic4Review or your generic imagery/detection layers in iD/JOSM and such).
1 Like
As a general rule of thumb you are correct because it is considered an automated edit but you still can make automated edits or imports under certain circumstances. It depends on the kind of data, whether the data has been (human) verified, is legal to import, and does not duplicate existing data. All automated edits need to be reviewed beforehand and consulted with the OSM admins.
If I am not mistaken Facebook has already pushed many AI detected road network automated edits in batches over the last decade or so of undeveloped areas, far east Asia, and Oceania. Often, this basic unlabeled road network has sparked local mapping activity in these areas. So, it depends. But, every batch had to be reviewed. This sort of thing is not feasible with Mapillary (yet).
I think the tool front has improved with Rapid and JOSM significantly over the last few years. Things are not perfect yet but especially Rapid seems to integrate Mapillary quite well. JOSM seems to lag a bit behind but it looks like the JOSM developers also have decided to ultimately make Mapillary a first class citizen (with traffic sign and point features vector tile Mapillary map layers, plus image preview 
) for mapping. There is also a Mapillary plug‑in for JOSM, which also integrates Mapillary with moderate success. So, I would say that the situation is far less bleak than just a few years ago.
1 Like
I might be out of the loop on this one, although I have generally kept track of where tools stand. I’m looking at Rapid and it doesn’t really seem to do anything that iD doesn’t already do, may be a slightly improved interface. I mentioned “quickly and efficiently”, which was the key point I wanted to highlight about the current tools. They are capable of using Mapillary, but they haven’t gone much beyond the basics - just showing the layers and preview. But they were capable of this for many years. If anything I would definitely say JOSM’s plugin is more capable here as you can, for example, actually right-click detected features to add them even if in a limited and somewhat cumbersome fashion. I guess I just wanted to say that my UX expectations are higher than current tooling. In any case, I also wouldn’t imply that the situation is “bleak” as such, just that significant improvement is still needed.
1 Like
I think I know what you mean; there is no one‑click‑to‑OSM verification tool for Mapillary detections. However, I am not sure this is ever going to work or is actually desired for OSM due to how OSM works and the database is structured. Furthermore, a significant number of detections would need repositioning or re‑tagging, so that one click would not suffice to speed up things. This is something that Bee Maps: A strong contender to Mapillary! does, where contributors can train the system and correct existing data points through basically one click verification. However, Bee Maps does not need to export their detections to an external structured database like OSM, so this makes their life a bit easier.