How is the number of attempted matches determined when no EXIF GPS?

I’m just testing a dataset obtained over the past 2 days with my phone with the latest native WebODM release, with no GPS data in EXIF.
When operating in a small area the coordinates are likely to be recorded as the same position in every image, leading to processing problems, as lat and long are only recorded to nearest arc second, or typically ~30m.

There are 374 images, and close to 70000 attempted matches, still in progress, so each image is checked against nearly 200 others (actually 186).

Is this a limit for attempted matches, or is it determined in some mysterious way?

1 Like

Yes, it uses bag-of-words for matching, so it essentially indexes and sorts features and uses that index to reduce the total number of matches to the likeliest without having to do a full pairwise matching.


Also, I have trouble not calling it barrel of words matching, which always invokes barrel of monkeys imagery and gets confused looks from real CS folk.


This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.