Ben’s 2D photo map tagger is a manual process that only needs one photo taken.
For a process like you list, one pixel at time, more automated approach, I’m working on that now, and didn’t post my first results, solely due to the results weren’t good (reflections from glass mostly the cause) and I haven’t had time to do it again yet. It worked, though, with 2 videos taken on my phone, using a phone browser. Adjusting the original code to fit the expected PB style of map would help, as I had to munge the results into PB format. Relatively easy, and I’ll do that shortly.