Thanks for the pointers everyone! I ended up putting...

Thanks for the pointers everyone!

I ended up putting together a thing that uses OpenAI's API with gpt-4o-mini to suggest OSM tags for a picture: https://github.com/iandees/image-to-osm

I'll try putting it up on the web somewhere if there's interest. I spent about 15¢ testing it a few dozen times:

A screenshot of a web app that shows an image of a storefront with some phone numbers and opening hours under the name of the place. Underneath, a text box with OpenStreetMap tags suggested, a warning to verify using the tags, and a button to copy them to the clipboard.

Like 15 August at 5:18 | Open on social.modest.com

4 comments

Ilya Zverev

@ian Wow! I just thought, adding a button to Every Door to use this would be sooo easy. But the multitude of edge cases makes me wary.

15 August at 8:37 | Open on en.osm.town

Ian Dees

@zverik yea this was inspired by me manually finding the hours on a place's website, copying the text into chatgpt, asking for the OSM format, and pasting it into every door. I thought why not cut out the middleman? 😁

15 August at 12:06 | Open on social.modest.com

Ian Dees

Give it a shot: https://image-to-osm.vercel.app/

Take a picture, get OSM tags. Put tags in mobile editor.

I'm still thinking about how to make the output useful. I don't want this data to go straight to OSM, but it sure makes some of the tedious part of OSM mapping while walking easier. Please give feedback or ideas!

15 August at 17:10 | Open on social.modest.com

Ian Dees

Based on some feedback I got from people that wanted to use it for more than just storefronts, I tweaked the prompt on my image-to-osm tool and it is happier to give OSM tags for generic stuff like bike parking and waste bins.

https://image-to-osm.vercel.app/

27 August at 4:45 | Open on social.modest.com

Go Up