this post was submitted on 06 Sep 2025
11 points (86.7% liked)

OpenStreetMap community

5195 readers
1 users here now

Everything #OpenStreetMap related is welcome: software releases, showing of your work, questions about how to tag something, as long as it has to do with OpenStreetMap or OpenStreetMap-related software.

OpenStreetMap is a map of the world, created by people like you and free to use under an open license.

Join OpenStreetMap and start mapping: https://www.openstreetmap.org/.

There are many communication channels about OSM, many organized around a certain country or region. Discover them on https://openstreetmap.community/

https://mapcomplete.org/ is an easy-to-use website to view, edit and add points (such as shops, restaurants and others)

https://learnosm.org/en/ has a lot of information for beginners too.

founded 5 years ago
MODERATORS
 

One thing that makes streetcomplete a bit of a chore for me is entering opening times: depending on where you are, you might have to specify it differently for every single day, resulting in a lot of typing. That is more time I stare at my phone and use my hands in an unergonomic position, leading to soreness in neck, hands etc.

So I was thinking: Surely we are technically able to extract those data from an image?

If I would have time to implement it (which unfortunately I won't I think) I'd just try to find a prompt that works well, send it to an LLM like Mistral (bc. in EU), get the info back, let it get checked by the user and then enter it to openstreetmaps. That would require only me to provide my api token, which I would be willing to pay for (assuming my estimation of this being in the cents volume are right).

Now:

  1. Does something like that exist and I just don't know about it?
  2. If somebody is willing to try that out I can provide some example data, collect more and test it.
top 2 comments
sorted by: hot top controversial new old
[โ€“] ame@breta.moe 2 points 2 months ago* (last edited 2 months ago) (1 children)

@DonnerWolfBach There were some experiements by @ian and @zverik with extracting osm tags from pictures using gpt-4o-mini in everydoor: https://en.osm.town/@everydoor/113192204761150958

The model is pretty out of date these days, so I imagine there's room for improvement!

#OpenStreetMap