Import based on similarity

[Gilles Caulier]

Exactly,

Import file and compute fingerprints in the database. At the end, run
the maintenance tool for the similarity stuff and all is ready for the
items reviewed by the user.

Best regards

Gilles

Le ven. 12 sept. 2025 à 12:08, Remco Viëtor <remco.vietor at wanadoo.fr> a écrit :
>
> On vendredi 12 septembre 2025 11:53:58 heure d’été d’Europe centrale Gilles
> Caulier wrote:
> > Le ven. 12 sept. 2025 à 11:18, Remco Viëtor <remco.vietor at wanadoo.fr> a
> écrit :
> > > On vendredi 12 septembre 2025 10:27:02 heure d’été d’Europe centrale
> > > Benoît
> > >
> > > Barbier wrote:
> > > > Hello,
> > > >
> > > > Is it possible to import based on similarity, or is it better to import
> > > > everything, then update the fingerprints and find duplicates based on
> > > > similarity?
> > > >
> > > >
> > > > Thanks,
> > > >
> > > >
> > > > --
> > > > Benoît
> > >
> > > I think the fingerprints are required to determine similarity. So doing a
> > > similarity test *before* import would mean reading all files, calculate
> > > and
> > > store (where?) the fingerprints, and compare against the database.
> > > Then on import you will have to read all the files *again*...
> > >
> > > I don't think digikam offers the possibility to calculate fingerprints for
> > > files not in its database (so not yet imported).
> >
> > Typically, the fingerprints must be calculated while downloading and
> > byte-array stored in the database, as with the similarity or the
> > maintenance tool.
> >
> > This means that all files are backported in the collection. As the
> > fingerprints are already in the database, the user must go to the
> > similarity tool to review the duplicates items as usual.
> >
> > Implementing this way is not too intrusive in source code. Tool to
> > compute the fingerprints on a set of images from the collection
> > already exists, It must be started at the end of downloading files
> > from the camera. This requires just a new checkbox in the Import tool
> > settings and the rules to start the job at the end of the import
> > process.
> >
> > Voilà for the concept.
> >
> > My best
> >
> > Gilles Caulier
>
> So, if I understand correctly, what you suggest is calculating the fingerprint
> on import, and then test for similarity. So it would still not be possible to
> decide which files to import based on similarity, i.e. get a similarity index
> *without* importing (which was basically what OP wanted)
>
> Remco
>
>
>