Website for photo album

Loving the shit out of this right now


Hosting about 11 gigs worth of photos right now (Mostly from my G Photos takeout) and it runs like a dream

p

EDIT: Screenie

5 Likes

If you don’t mind, what hardware specs are you running it on? Curious how hungry it is.

We recommend hosting PhotoPrism on a server with at least 2 cores and 4 GB of memory . Beyond these minimum requirements, the amount of RAM should match the number of cores. Indexing large photo and video collections significantly benefits from fast, local SSD storage.

Source: GitHub - photoprism/photoprism: Personal Photo Management powered by Go and Google TensorFlow

1 Like

They even have installation instructions for Raspberry Pi:

1 Like

Right, I read the directions but I wondered if it actually uses that much or if that’s just their recommendation because memory is cheap and some people are going to put ten thousand 4k pictures on there. I have an ability to get first hand information from @thagoat so why not ask?

1 Like

1 Like

It’s running on a Debian vm, 4 core Xeon 2.5 with 6 G of ram. So, beefy enough to run the docker container plus ghost, miniflux and watchtower.

1 Like

Incredibly fast.

1 Like

So only using a fraction of the “minimum” 4 GB RAM? :sweat_smile:

At rest, yes. Memory usage is high when importing photos. That’s probably why the “requirement” is 4 gigs.

1 Like

Probably uses a fair bit when classifying/tagging images using ML (Tensor flow) as well. Sounds like it includes a trained model (which is the most resource intensive part of ML), but it’ll still have to do a bit of work to process and classify each image.

2 Likes

giphy

1 Like

Processing is very slow… I have 14300 pictures to process, plus movies.

Side note - I quote this all the time to my wife. His voice, shirt removal, and all. She thinks I’m weird as fuck, but I think it’s hilarious. Early Sandler movies are the fucking best. Happy Gilmore, Billy Madison, Waterboy, all so damn good.

2 Likes

Main feature lacking in PhotoPrism compared to Google/Apple Photos is face recognition etc? Can’t search for “Joe” or “dog” or whatever? :slight_smile:

I’m comparing two systems at the moment: PhotoPrism and PhotoStructure. Currently, my impression is that PhotoPrism is more feature-complete (it has search, ML-based tagging, etc) whereas PhotoStructure is more polished and has way more advanced deduplication and detection of time the photo was captured. These heuristics are more advanced than I’ve seen in any other self-hosted photo app.

I think I’m going to stick with PhotoStructure. There’s some missing features, but the developer is working on them, and I think over time it’ll catch up to PhotoPrism in terms of features and beat it in terms of quality. I’ve been pretty active on their forum (I’m also “Daniel” there) and developer is extremely helpful and responsive. My understanding with PhotoStructure is that its developer is working on it full-time, and he’s going to add some paid plans with advanced features to fund the development, which seems like a reasonable approach to me.

If you used Google Photos previously, PhotoStructure has support for importing metadata from Google Photos. The Google “Takeout” archive with all your Google Photos has some extra JSON files with all the face tags that Google automatically added. PhotoStructure can ingest all these and will display all the same tags in its upcoming release :slight_smile:

Face recognition is coming to PhotoStructure in the upcoming release! :slight_smile:

2 Likes

Oh, also, PhotoStructure has both a desktop app and a Docker container, so you can run it on your desktop PC, do the initial import there, then rsync the entire library to your server. If you want to test it out, you can just download the desktop app and run it :slight_smile:

I picked up a storage VPS (2 TB space) and a regular VPS (16 GB RAM, 80 GB NVMe disk, 3 vCores with 1 dedicated) from HostHatch during their Black Friday sales, and I’m planning on running it there (share the space with NFS, and run it on the NVMe VPS).

1 Like