Hi, We have another open source solution for you - Photonix (https://photonix.org/). We've been working on it for a few years now but it's getting more feature-rich every month.
Key features: web-based, ML auto-detection of objects, colors and styles, map view, Android and iOS apps, ARM/Raspberry Pi support, works with your existing photo folder structure.
Face recognition/matching is almost done and we're also planning a cloud-hosted service for those that don't want to/can't self-host.
For those interested in the tech - the backend is Python, frontend is React-based, ML is done with Tensorflow/Keras, building is done via Docker, APIs use GraphQL.
Hi @rochak. Thanks for the feedback. We do allow for tagging but the UI is a bit hidden on the photo detail screen right now. If you click a thumbnail and then scroll down you can see the auto generated tags, histogram, location and can add tags of your own. I'm going to change that part of the UI soon so it comes in from the right I think and has an icon to access it.
Hi @rochak. I've been wanting something similar for a while and have been working on my own solution with a focus on privacy and open source. For this reason all image analysis (AI/location awareness etc.) is done locally and it can all run within a Docker container. This is the site so far and there's also an online demo: https://photonix.org/
For those that don't know yet, Photonix is a self-hosted photo management web app with machine learning image analysis (object detection, style recognition, color identification etc.). I've been working on it for a couple of years now but development pace has really picked up since I hired another developer work with me.
I'm aiming to give monthly progress from now on so this blog post details all the new features that have been added in the last month-and-a-bit. This time we have star ratings, range filtering, text search, performance optimisations, general tagging and more.
Please give it a try if you have time - we have a demo server which makes it very simple. I'd love to hear your thoughts. Thank you for your time, Damian.
You might just want to try making a copy of a smaller selection of your photo library in a new folder and giving that to Photonix to try out. That way it doesn't matter if anything were to get removed.
Yep, it should work fine with a read-only photo volume as it doesn't make any changes to photo files by default - it just needs to write to the DB which is configured as another container in the Docker Compose file.
The thinking behind GraphQL was to allow for advanced filtering, supporting all the attributes we store without a lot of extra API work. The GraphiQL web interface makes it quite nice to explore the data and is bundled in and accessible at /graphql . GraphQL also has "subscriptions" which allows for pushing data from the server. The Apollo JS library I used also provides built-in extras like caching.
At the moment I create "thumbnails" at various sizes on discovery (4k, 2k, etc). Currently the 4k version is displayed but this will adapt and download the smaller versions on smaller devices.
The UI for metadata is a bit unintuitive at the moment but you can scroll down to see it when you're viewing the fullscreen photos.
I definitely hear you regarding duplication. This whole project evolved from a script I wrote to do just that. There is a concept of multiple versions (files) of an image so different edits with the same metadata timestamp will show only once and you'll be able to select the preferred one. I prefer to handle things this way to start and then suggest deletions based on same hash rather than deleting up front.
I don't have any experience of Lightroom but I can have a go at reading the files if you think it's useful.
Thanks for the feedback. I agree, there should some progress display while downloading. I'll be sure to make sure that's there. There are other optimisations I want to make in this area like selecting a resized version to download depending on the current screen dimensions and pixel density. This will improve loading times.
Installation is fairly simple with Docker, frontend is web-based (React), backend is Python with a sprinkling of Tensorflow. So far auto-tagging of photos by location, object detection and colour is fairly decent. UI is progressing and is useable on most devices, though quite minimal.
Please feel free to check out the demo site and the GitHub issues. I'd really appreciate feedback and help. Thanks.
Key features: web-based, ML auto-detection of objects, colors and styles, map view, Android and iOS apps, ARM/Raspberry Pi support, works with your existing photo folder structure.
Face recognition/matching is almost done and we're also planning a cloud-hosted service for those that don't want to/can't self-host.
For those interested in the tech - the backend is Python, frontend is React-based, ML is done with Tensorflow/Keras, building is done via Docker, APIs use GraphQL.
Hopefully see some of you over at https://github.com/photonixapp/photonix