Closed pierotofy closed 3 months ago
What sizes do you want / need tested?
A variety would be best (small, medium, large), but the main concern is just trying to answer the bullet points I mentioned above.
Cool. I've got it running on a 1900 dataset. Seems like a decent size (larger than most test datasets) without taking overly long to run (should be done tomorrow afternoon). All defaults, one running on current and one on a fresh build of this branch.
So far, I don't think I see any computational time differences in OpenSfM:
Only difference I see so far is progress bar differences for the same stage:
Definitely got some weird errors in the northeastern corner:
Kakuma_17A_Flight_02-report.pdf Kakuma_17A_Flight_02-trackdec-report.pdf
trackdec left, current right
This PR brings the changes suggested from https://github.com/mapillary/OpenSfM/pull/641 into ODM.
On small datasets this seems to decrease relative accuracy, so perhaps this might need to be adjusted to only be used on larger datasets (how large?)
Please help test, in particular: