I’ve been using mapwith.ai, RapiD editor, and enabled the msBuildings vector layer. in rural areas it’s been quite nice, in thick built up urban areas much less.
my workflow has been first mapping with the RapiD editor, then open the same area with JOSM, and review and refine, like in this example, reviewed here.
the obvious question is can we use the msBuildings vector layer straight into JOSM? and which would be the correct workflow in this setup?
The answer to your question is on the Mapwith.ai Wiki FAQ:
7. Will you offer JOSM support?
Yes, we have a JOSM plugin! GitHub - KaartGroup/JOSM_MapWIthAI_plugin: A fork of JOSM/MapWithAI
to simplify the answer: start JOSM, open the Editing Preferences window, Plugins tab, then start typing ‘mapwithai’, before you’re done typing, you’ll see the plugin you want (or the -dev version).
JOSM MapWithAI plugin maintainer here:
I’m planning on removing the mapwithai-dev version soonish. I’ve recently migrated the repo to the JOSM GitHub group (GitHub - JOSM/MapWithAI), and I probably won’t be able to update the -dev version easily for awhile, as I’m moving away from using GitLab for my development/release process.
I am also willing to listen to feedback, so if you find something confusing, please let me know (open a ticket on JOSM Trac). There are some things that I know confuse users, but which were put into place to avoid imports.
Prior to load footprints in JOSM, I used to filter candidates in Qgis. That’s mainly because you may need conflation, so every candidate that touches OSM buildings should not be imported.
Once you have a candidate layer, you should check alignment. I used Strava to align highways THEN used them as a reference to fix possible general buildings offset.
Last, using keyboard shortcuts, I fixed many buildings that needed rotations, say 30-40% of them. I guess rotation issue is not related to projection applied somewhere, but due to MS simplification algorithm.