It seems there’re many people looking for a way to use Seafile as a self-hosted Google photos / Immich. I believe most functionalities are already there, it just needs polished mobile Drive clients to turn Seafile into a killer app for managing photos and videos!! And AI facial recognition/tagging would be very cool too.
Here’re related posts:
/t/photos-gallery-timeline/11374
/t/mobile-app-adding-delete-after-upload-on-picture-video-files/23149
/t/seafile-video-player/22636
/t/can-this-replace-google-drive-photos/13956
/t/provide-a-thumbnailing-api/22842
1 Like
AI facial recognition is already added in version 13.0 (Seafile AI extension - Seafile Admin Manual)
Regarding the posts:
Here’re related posts:
(done) /t/photos-gallery-timeline/11374
/t/mobile-app-adding-delete-after-upload-on-picture-video-files/23149
(done) /t/seafile-video-player/22636
(done) /t/can-this-replace-google-drive-photos/13956
(done) /t/provide-a-thumbnailing-api/22842
So, in summary, I think most features are already implemented in version 13.0. Just give it a try and let me know what you are missing.
2 Likes
Wowww. I just tested Seafile Server 13. Most features are indeed there to be a photo/video backup solution!
- the facial tagging doesn’t work well though.
- videos dont have a thumbnail (without Thumbnail Server) and people tagging
- the mobile client needs more polishing like people tagging and world map/location grouping.
You can follow this document Thumbnail Server - Seafile Admin Manual to add thumbnail server and enable thumbnail for videos.
- the facial tagging doesn’t work well though.
Can you describe more on this? The facial recognition works well in my case.
- the mobile client needs more polishing like people tagging and world map/location grouping.
We will check this.
1 Like
This is what I got from facial tagging. Basically it doesn’t recognize anyone and just group everybody as “Unknown people”. I am testing Server Version: 13.0.11 with cpu face tagging. Btw, I dont have LLM configured. Is that necessary? I put these in the .env:
ENABLE_SEAFILE_AI=true
SEAFILE_AI_LLM_TYPE=other
SEAFILE_AI_LLM_URL=https://api.openai.com/v1 # your LLM API endpoint
SEAFILE_AI_LLM_KEY=<your LLM access key>
SEAFILE_AI_LLM_MODEL=gpt-4o-mini # your model id
The pictures are just random download pictures from the net. There’re many people in the pictures. I’ve blurred their faces in this screenshot to protect their privacy. But the pictures in the library are not blurred.
Oh, and to make this really rock as a photo/video solution. It should be very easy to access these features. For example, add more Workspace for:
- Photos
- Videos
- People
- Map / Places
Face recognition feature does not require LLM configured.
It is recommended to use over 100 images of real people for testing, with each image containing one or two individuals.
The algorithm will cluster the individuals detected in the images every 24 hours. If you provide two few images, the algorithm will not be successful.
I’ve uploaded more than 400 pictures and left the server on for more than 24 hours. Still, they’re all classified as “Unknown people”. Also, the Gallery view and people tagging don’t seem to support pictures in .avif nor .heic/.heif format.
Do you have more than one “Unknown people” category such as the screenshot below?
These formats are not supported right now.
