Is there consensus on mapping pavements (sidewalks) separately to roads?

[Matija_Nalis]

@Minh_Nguyen

If the goal is to maximize coverage while prioritizing ease of entry, then I wonder what folks in this thread think of using automated approaches based on imports, road geometry heuristics, or computer vision as a starting point. None of these approaches is perfect by any means, but the quality triangle only has so many vertices.

I doubt that bringing “do you like AI in OSM?” into any topic will make it less contentious

• I have no problem (and in fact, encourage if they like doing that!) human mappers to convert sidewalk:left=yes or cycleway:right=track into separate ways tagged as highway=footway/cycleway in order to improve details and add extra details (surface, smoothness, lit, barrier, obstacle, …)
  (though they should remember to replace original value with *=separate in order to avoid confusion)
• I have no problem with automated imports and AI adding data where no data is existing, providing they add the tags to indicate it was result of import / AI detection.
• I am even OK with some kind of program or AI assisting human in tedious task (like making sure drawn highway=footway is parallel to original road, or follows aerial imagery) as long as the human does the important part of the work (just in less clicks)
• I am however not OK with AI / import tools messing autonomously in areas where human mappers are doing that same type of work. I don’t find them nowhere near good enough for that, and they damage the community (not only by dubious data quality, but on sociological level too - see several articles linked at bottom of Import - OpenStreetMap Wiki). IMHO, that amount of bot autonomy should be accepted only when we’re at level of removing humans completely from OSM mapping, and having autonomous drones map everything instead. However, I’m not sure if our AI overlords would care much about OSM at that point