I'm bumping this, just read through older related posts, wish to revisit.
I'm presently working a complex scene, only some 1500 photos in a cave chamber, but the complexities, e.g. sloped mud banks, 25' domes, highly occluded huge cobble-lined floor, multiple ceilings... challenge capture to meet the ideal, all images aligning out of the box. I've gradually managed to get this room to come together, so the photography obviously provides adequate overlap and overall coverage. The obvious upshot of all this is the splintering into a dozen and more small components with 2-8 images, with five or more components 20-300. I'm now up to 1200 in a component, but not without significant slow down because of so many smaller components to either delete or, in an alternate workflow, isolating the larger components by deactivating all photos, then only enabling cameras in a large one alongside whatever handful of sequential/overlapping pairs being aligned with CPs.
Naturally, when I have two larger components I'm ready to relate I first try setting Merge components only to True and Force component rematch to True. The components either merge or don't, but either way, no smaller components to deal with. To note, I also set all inputs to Use all image features. When this lean approach fails, and when down the laborious road a merged component confirms the dataset supports alignment, this opens to question what explains it taking so much time? I submit it's a combination of the inherent complexity of so many variables that a yet savvy user struggles against whittling at the problem, that a going strategy is also thwarted by clunky workflow. Some of this relates to there being no slick functionality in how RC supports project hygiene, another factor, one relating to project hygiene, is how older components influence younger components and in turn new alignments.
The moment you resort to setting Force component rematch to False or Merge components only to False, you may have an older component with more poses but containing drift, a younger component that splits the older, but now merges added geometry. Since you don't want to lose what you have in the older component, you keep it, but now you've invited a trap. There may be deactivated images from among so many smaller components that are needed to tie it all together, but which ones? You hunt for related images, activate them, add CPs, and if you select the right combination of alignment settings and which images to activate, you might get something, but it's far too easy to create a bigger mess while you're juggling so many alignment settings, activation states, and struggling with a mental map of which photos hold promise, where do they live, etc. etc.
Ideally, RC doesn't make complex projects more complex by forcing the user to think unnecessarily hard about workflow. It's our job to understand the tools and how this all works, but we only have so many bullets to spend on studying tiny features, tiny alignment settings (please, another feature request I've begged for, a zoom function with the scroll wheel to increase font size in the text panes!). So, how best to streamline project hygiene? If RC generated a folder per alignment, allowed one to pull selected components out of the folder before deleting the folder, that certainly would go far to keep the project uncluttered, while keeping younger components from unwittingly upsetting newer alignments.
Please sign in to leave a comment.