Workflow for similar images

[mowgli]

Hello,

it would be nice to have a semi automatic workflow for similar images (d).

If similarity = 100% then move the bigger (size) image to... end If

If similarity = 100% then move image with "STRING" in the name to... end If

If similarity = 100% then move the bigger (size and resolution) image to... end If

If similarity <= 95% then move the bigger image to... end If

Do you understand? That will be !!!GREAT!!!

And an export of the similarlist to a textfile.

Reported by: *anonymous

[mowgli]

Logged In: YES user_id=1986130 Originator: NO

Thanks for the idea!

This workflow is cleanly a task for external script. My original idea was that Geeqie will write everything to image metadata and the external command can operate on them. The similarity value is however a bit different. It does not probably make sense to save it to the metadata.

Other option is to pass it to the external command on commandline. This will need an extension to the commandline format.

Original comment by: nadvornik

[mowgli]

Logged In: NO

Hello,

changing the Metadata is no good idea. Because the file has a new checksum (md5/sha1...). The checksum is the best identifier for a file. I have a big collection of images (70.000), the mainkey are the checksums in a database.

I think, the best way is an extern command with basic build in scripts, and the possibility to define selfmade scripts (bash?).

Now I use the "libpuzzle"[1] and a complex bash script behind with "dialog" to define the sort rules. But the similaritysearch is not so flexible as gqview. If I could the results from geeqie to a textfile, I can use my script resume

http://libpuzzle.pureftpd.org/project/libpuzzle

Original comment by: nobody

[mowgli]

Hmmm... just leave my idea about this. How about using something like lua as an internal script engine? This would allow to make it platform independent.

Original comment by: mowgli

[mowgli]

Would be an idea for lua integration.