[digiKam-users] Imilarity.db

Boz Avram theodormcl at gmail.com
Fri Sep 3 17:14:11 BST 2021


Hello,

After further investigation, I can confirm that the two fields, imageid1 and imageid2 of the ImageSimilarity table of the Similarity.db indeed point to appropriate images within digikam4.db tables (specifically, Images table; need also to use Albums and AlbumsRoots tables to get to a directory where the pics of interest are). 

Btw, as everyone knows, DigiKam takes a long time to compute similarities, but on my collection the quality of the similarity results is vastly better than the results I was getting from a couple of commercial standalone programs I tried because the latter were much faster in arriving at the results. 

The “issue” I was originally inquiring about most likely came due to a transient state of the digikam4.db due to my move from local storage to the same Albums structure on a NAS. —Theo

Sent from my iPad

> On Sep 1, 2021, at 9:15 AM, Boz Avram <theodormcl at gmail.com> wrote:
> 
> 
> Hello,
> First, a big thank you to the developers (and contributing users) for providing us an ever improving, and already a very capable package.
> 
> My question is regarding the similarity db, with an apriori apology  if the answer was provided somewhere (i tried, really tried to find it, but did not have much success):
> Using DB Browser for SQlite, i see a table ImageSimilarity, which has two fields of interest to me, Imageid1 and Imageid2 (along with a field that shows computed similarity between the two. If, for one record, I use the two image IDs to lookup in digikam4.db, I don’t get images that have any similarity. Hence the ImageSimilarity table IDs seem to point to something else? 
> Any clarification of this similarity DB, which would allow me to manually locate the  images of similar pairs would be highly appreciated. 
> The context of this inquiry is that I would like to programmatically remove some of the duplicates, as there are thousands of them.
> Thanks much in advance, Theo
> 
> Sent from my iPad


More information about the Digikam-users mailing list