https://bugs.kde.org/show_bug.cgi?id=479090
Bug ID: 479090
Summary: [Request (Plugin)] Object Recognition for automatic
tag generation, easier searching
Classification: Applications
Product: digikam
Version: unspecified
Platform: Other
OS: Other
Status: REPORTED
Severity: wishlist
Priority: NOR
Component: general
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---
I was searching for an easy way to locate pictures within my collection when
only knowing the content of the picture, and i found the Recognize Anything
Paper (https://arxiv.org/pdf/2306.03514.pdf) and connected Github
(https://github.com/xinyu1205/recognize-anything) from the OPPO Research Team
licensed under the Apache-2.0 license. (But there are other open source
solutions that would work too)
This would allow for something analogous to the Face Tags, but more
generalised, to enable easy searching. Just by typing in the subject you can
find pictures the AI tagged, within your entire collection.
For good integration you would need digikam to track these "Machine Tags" (in a
separate .db, similar to recognition.db) and allow for searching- which seems
easy enough since you can already search for tags with "A tag of the item
contains", so adding some UI elements to differentiate them from normal tags
and keep them separate for only search purposes would probably suffice.
In my opinion, you could add a popup just like face recognition to download the
python project, then pass pictures to the program and record the Machine Tags
in a separate .db file, which then can be used for searching.
Currently i am experimenting with adding these tags to the exif metadata, but
this is suboptimal for multiple reasons, and an integration would obviously be
superior.
--
You are receiving this mail because:
You are watching all bug changes.
