github imagegenius/docker-immich v1.48.0-ig55

20 months ago

ImageGenius Changes:

use machine-learning with python

immich Changes:

February 2023 (version V1.48.0)

🔴 Breaking Changes and Important Notes for this release 🔴

  • The server and the mobile app must be on the same version to function correctly.
  • There are changes to the docker-compose.yml file for the machine-learning portion, and include the model-cache volume at the end of the file as shown here. Please update yours accordingly.
  • Temporarily dropping Machine Learning support for Arm64 and ArmV7 due to PyTorch not supporting these architectures from the Python stable wheels

Highlights

Thanks to our talented contributors, I am excited to bring you to one of the biggest releases yet of Immich that I hope you will enjoy. This release brings a lot of new features and improvements to the app. Here are some of the highlights:

  • iOS Background Uploads
  • Machine Learning moved to Python-based Image
  • Auto-link Live Photos

iOS Background Uploads

I want to thank @martyfuhry for diving into this feature and his wife for wanting the feature :P.

iOS users can now upload photos and videos in the background. This means that you can take photos and videos, which will be uploaded to your server without opening the app.
The background mechanism from the iPhone has two modes, a shorter task which runs around every 10 minutes and lasts for about 30 seconds, and a longer task which runs every 30-45 minutes and lasts longer. From our extensive testing, the app can upload about 1000 pictures overnight without plugging in and on WIFI.

Please enable Background App Refresh in Settings > General > Background App Refresh for this feature to work.
You can let the background uploads run on WIFI only or both WIFI and Cellular. This can be configured in the same menu in the abovementioned settings.

You can get notified when the background task is running by enabling the notification setting from the Settings menu on Immich mobile app.

We are looking for feedback and bug reports on this feature. Please let us know if you have any issues with it.

Machine Learning moved to Python-based Image

We are planning to revamp our search capability. The first step to making it smarter is to move our machine learning project from NodeJs based to Python based. This will allow us to use more powerful machine-learning libraries and improve the search.

We are planning to add facial recognition in the near future and also use SBERT for image search. This will allow you to search for images based on the content of the image.

image

This is a breaking change to the current setup. Please update your docker-compose.yml file for the machine-learning container accordingly. We are also dropping ML support for Arm64 and ArmV7 for now due to PyTorch not supporting these architectures from the Python stable wheels. We are looking into ways to support these architectures in the future.

Auto-link Live Photos

LivePhotos’ motion and still parts are now automatically linked together. This means that you can upload two separate files from the web interface or from the CLI, and they will be linked into one LivePhoto asset and displayed correctly on the timeline.

For previously uploaded motion and picture parts of the LivePhoto asset from the web or the CLI, you can re-run the metadata extraction job to link them together.


And like other releases, we have made many other improvements and bug fixes for better performance and UI consistency across platforms. Please take a look at the full changelog below.

As always, thank you for using Immich, and I hope you enjoy this release.

Remember to support the project!


🎉 Cheer! 🎉

Support

If you find the project helpful and it helps you in some ways, you can support the project one time or monthly from GitHub Sponsors

It is a great way to let me know that you want me to continue developing and working on this project for years to come.

What's Changed

Server

Mobile

Web

Others

New Contributors

Full Changelog: immich-app/immich@v1.47.3...v1.48

Don't miss a new docker-immich release

NewReleases is sending notifications on new releases.