But that means that the absence of a tree structure is in the way you use the system, not in the system itself. Having collections of images allows digikam to check for new images on start-up, without having to trawl the whole HDD. It’s slow enough as it is on a few thousand images in one tree.
I wonder if you could get away with just three collections:
- finished images
- images “under treatment”
- everything else.
And I can’t help wondering why you need 17 versions of an image… You could also use other means to reduce the number of results: stars, colour labels, and pick labels. And images can be grouped, and only the group head shown (not sure what that does with searches for tagged images, though).