Hi all,
I’m new to mapping and getting going in Yonkers (Westchester). Buildings are mostly missing here, so I want to start adding them from the Microsoft US Building Footprints, which are ODbL and line up well with Esri imagery. My plan is to work one neighborhood at a time in RapiD, checking each building against imagery before I add it and letting RapiD flag anything that overlaps what’s already there. Footprints come in as building=yes. I’ll square up or skip the ones that don’t match.
If I go past a neighborhood or two, I’ll write up a proper import page and use a separate account first. Wanted to say hi before I start. Anything local I should know, or is someone already on this?
My personal impression is that it’s still too much work to fix up the AI buildings rather than draw them myself.
Intuitively I would imagine an AI should be able to do a 99% perfect job of making all rectangular, properly cut out buildings. Make a row of houses look like an actual row (same shape, same size and parallel). However, in practice I end up spending so much time correcting the AI buildings that it’s just faster (and higher quality at the same time) to draw them with JOSM building plugin. I consult the ML buildings in rapid afterwards to double check whether I forgot something.
My suggestion would be to try each approach for half an hour and then count which method is actually faster. Curious to hear what you find.