colmap
delete image while reconstructing
Describe the bug It is not a real bug but I am observing colmap during the reconstruction of a huge dataset (about 6000 vertical images of a lanscape). At a certain point, few images were wrongly aligned and I was wondering if I can delete them and if it makes sense to do it during processing or at the end of it. Indeed, since only 1600 images have been aligned so far (as I said, processing is still going on), I am afraid that more images may match the 20ish images which are wrongly positioned, thus compromising the entire reconstruction. I see that if I double click on an image, a window pops-up with some information and I can delete the image. Will this affect the ongoing reconstruction or should I just wait for it to finish and then "clean" the sparse cloud? And if I delete these images, will they be used again later (or how can I force a second trial at the end of the processing)?
Expected behavior I would expect that deleting an image during sparse reconstruction, will make the following processing more stable.
Screenshots
In the image you see a sort of flat layer of gray points (sparse reconstruction of piece of landscape) and a few points going upwards, representing the wrongly positioned camera I would like to get rid of. (please do not consider the cameras -red shapes- since they are in perspective).
The simplest solution may be:
Run on all images -> Record wrong registrated images -> Filter them out-> Run COLMAP again on filtered images.
If you don't want to run colmap twice, you can also delete the wrong images in the colmap bin files by hand, and then slightly refine the reconstruction. It would be easier to do this by pycolmap
