Can I use this space to comment on how amazing Immich is? I self host lots of stuff, and there’s this one tier above everything else that’s currently, and exclusively, held by Home Assistant and Immich. It is actually _better_ than Google photos (if you keep your db and thumbs on ssd, and run the top model for image search). You give up nothing, and own all your data.
I migrated over from google photos 2 years ago. It has been nothing but amazing. No wonder google has it in its crosshairs.
Don't they block NextCloud sync in the Play Store, for similar reasons?
yeah same, I'm in the process of migrating so I have both google photo and immich, and honestly immich is just as good.
I actually find the semantic search of immich slightly better.
What model do you recommend for image search?
Not OP, but CLIP from OpenAi (2021) seems pretty standard and gives great results at least in English (not so good in rarer languages).
https://opencv.org/blog/clip/
Essentially CLIP lets to encode both text and images in same vector space.
It is really easy and pretty fast too generate embeddings. Took less than hour on Google Colab.
I made a quick and dirty Flask app that lets me query my own collection of pictures and provide most relevant ones via cosine similarity.
You can query pretty much anything on CLIP (metaphors, lightning, object, time, location etc).
From what I understand many photo apps offer CLIP embedding search these days including Immich - https://meichthys.github.io/foss_photo_libraries/
Alternatives could be something like BLIP.
This is what I use:
ViT-SO400M-16-SigLIP2-384__webli
I think I found it because it was recommended by Immich as the best, but it still only took a day or two to run against my 5 thousand assets. I’ve tested it against whatever Google is using (I keep a part of my library on Google Photos), and it’s far better.