At least that's what I do... if you don't know the person, why keep the face? You're not winning a price for having the most random faces in your library 😜
It was not clear to me what the cross does. From the docs I understood it was only to remove a "face" that was not really a face.
Maybe it should be mentioned that you can use this also to remove faces you don't want to keep
Ideed, it also works for faces that are a face but you don't care about (we wrote the docs when we were happy it detects faces at all)
If you like, you can improve our docs by simply pressing the edit button in the upper right corner and then editing the markdown file (if you know how to format documents with markdown)
1 reply
howdy folks, there are several large videos in my photoprism originals directory, I've determined the ones I want to inspect with
ncdu but I want to check their contents. Is there a way to do this in photoprism? I don't get any results when searching for their filename in the GUI, despite the documentation saying that this should be possible https://docs.photoprism.org/user-guide/organize/search/ , all the filenames all look like 20190123_153450_E2E6B365.mp4 so the filename itself isn't helpful either.
6 replies
where do "keywords" originate? the exif data in the image has no keywords field, but there it is with correct keywords for the image
the labels are blank
From different sources such as color detection, location data, and filenames. And of course Exif and XMP if available.
my loose plan was to use an object detection api to update exif data in the "keyword" field
where objects detected are added to keywords
and then let photoprism do its thing
Well, you could search for the keywords but actual object detection includes knowing the position on the image. A keyword doesn't tell you this.
and then use photoprism to find all the keywords/tags
such as to find all "blue, subaru" pictures, etc
Labels might be even better for this
They can have a cover image, category and are visible on the Labels page (see FAQ)
i'd like it portable so in 10 years i can still find my images
Files are not modified at all right now, it's planned for later as there's always the risk of corrupting them. The YAML sidecar file is updated currently. If there is a label with the exact same name as a keyword, the photo is added to it automatically.
Anyone know how to set up library indexing on a set interval? Somebody apparently doesn't want to use WebDAV for importing new photos. Instead they're just adding them to the originals folder
Any job runner will do, make sure to index ideally while there's no activity and files don't change continuously. For example at night.
Here's an example config: https://dl.photoprism.org/docker/scheduler/
[job-exec "photoprism index"]
schedule = 0 0 1 * * *
container = photoprism
command = photoprism index --cleanup
no-overlap = true
I'm enjoying photoprism more and more, rediscovering some great memories while getting my photo collection organised and this wouldn't have happened without it, thank you so much for your work ❤
Glad to hear that! Seems we're getting closer to our vision 🙏
Flagged a few more more issues with "Help Wanted", e.g. because we don't have the right hardware to reproduce them: https://github.com/photoprism/photoprism/issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22
I could not find any detailed information o the privacy implications of using the google tensorflow for face recognition.
So how invasive is it?
1 reply
As far as I know, there’s a trained neural network that is part of PP. It analyzes the photos but does not communicate with the outside world in any way. Consider it a local helper program to classify photos or recognize faces.
Everything related to analyzing image contents happens on your own computer or server 👍 only maps are loaded from a cloud server as they are too large and need to be updated frequently
We'll update our docs / FAQs to make this more clear... when people read "Google TensorFlow" they may be confused indeed.
Which logo version do you prefer? Left or right?
I‘d even go as far as saying the bright neon prism/triangle would make a great icon 🙃
Probably too much for most, not sure if and where we can use it.
Its colors also bleed on many sites due to compression artefacts. JPEG aggressively removes details from dark images.
In case you haven't seen it yet: https://twitter.com/photoprism_app/status/1453680169529618446?t=xVvb4J8ABVCLnfzeXYPtvQ&s=19