[Support] kilrah/nextcloud-ffmpeg with pdlib

[ados]

13 hours ago, Kilrah said:

PDLib should now be available with the latest update.

Thank you for adding into your image.

Sadly getting the same issues when I installed manually.

You also need to load the module using the PHP file which is missing from the original image and thus most forks.

When you run the setup it isn't loaded.

 

This is when trying to access the Face Recognition option in administration settings.

 

You can confirm with command: php -m | grep pdlib

This will be blank as the module isn't loaded.

You need to add the following to the php.ini file in accordance with their documentation:

[pdlib]
extension="pdlib.so"

 

You can locate the expected path using: php -i |grep php.ini

There will be no php.ini file and making one causes the image to not start.

I think once that file is sorted it will load, any ideas?

Edited June 22 by ados

[Kilrah]

Typo in build script had caused the "latest" tag to not be updated. Fixed now, but it's usually recommended to use major version tags (:28, :29,...) on nextcloud to avoid nasty surprises on major version jumps. 

 

Edited June 22 by Kilrah

[ados]

18 hours ago, Kilrah said:

Typo in build script had caused the "latest" tag to not be updated. Fixed now, but it's usually recommended to use major version tags (:28, :29,...) on nextcloud to avoid nasty surprises on major version jumps. 

 

That resolved it, now loads PDLIB and FR works.

To assist those setting this up here is the short steps.

 

1. Once using the latest image open CLI on the host

 

2. Run the following: docker exec -it -u www-data Nextcloud ./occ face:setup --memory 5 GB --model 4

Assign the amount of RAM your willing/able to for image processing.

If your wanting to process higher resolution files especially HEIC than 2 GB+.

 

3. You also need set the PHP memory limit to equal or higher than above. You can do this with config files or a variable which will allow easy changing if needed. PHP_MEMORY_LIMIT and the value is in MB i.e. 5000M

 

4. Login to your NC and enable all the Memories app settings you want and do the same for Recognise.

Don't forget to enable "Index all media automatically (recommended)" in Memories.
Face Recognition settings can be adjusted later if your not happy with the results.

 

5. We need to setup Recognise, run the following using host CLI.

docker exec -it -u www-data Nextcloud ./occ recognize:download-models

docker exec -it -u www-data Nextcloud ./occ recognize:classify

 

6. We need to scan the library manually once and Memories will handle new files from here.

Again using CLI on the host run: docker exec -it -u www-data Nextcloud ./occ memories:index

 

7. Open the Memories app as the user (top bar) and setup your root paths.

These are areas you want Memories to process, allows you to exclude specific files if needed.

 

8. Enable face recognition for your user account (per user) in the user settings.

 

Things to note and general tips.

Make sure you're using CRON or Recognize won't work, find the radio button in basic settings.

To minimise HDD spin ensure your /data directory is your cache SSD.

The file data /var/www/html/data goes to the array of HDDs.

Ensure you have you have configured your Nextcloud config file with ffmpeg and best practice settings.

You can see any missconfigured best practice settings from the overview tab in admin settings.

Edited June 23 by ados