Support for `--folder` and `--user` parameters in `occ recognize:classify`

[wonx]

Describe the feature you'd like to request

I use nextcloud together with memories to browse my pictures. It integrates Recognize as part of its features to tag the content of the pictures themselves.

I have a collection of hundreds of thousands of pictures on network drives, which are accessed via nextcloud's external storage. Moreover, I have several test users that have access to those files, but don't use them actively, so the whole picture library is actually duplicated several times in nextcloud database.

So, when running a full indexing of my library, something I'm aware it can take several days, I'd like to skip those users and folders that I know don't contain any pictures or won't be used at all.

Describe the solution you'd like

I would love if there was some parameter in the occ recognize:classify to specify for which user I'd like to run the classificator, or limit it to a specific folder subtree. For instance:

occ recognize:classify --user myuser --folder "Media/Pictures"

This would limit the classification to my user, and circumscribed to my picture folder only (I don't need recognize to scan all my documents or work files). And would save me several days of processing time.

Describe alternatives you've considered

The only alternative would be to set up a separate nextcloud installation with only that user and an external storage pointing to the specific path of the picture library. But it's not realistic.

[jirislav]

This works for me: https://github.com/nextcloud/recognize/pull/1225

Feel free to adjust anything you need, I'll have a limited availability.

[joshtrichards]

Related:

• #785
• #625

[luxzg]

Any solution to force a scan to user / folder would be very helpful.

Few usage cases:

• manually import a new image archive folder from some external drive/location, run the commands only on that folder
• user uploads images from phone to a dedicated directory, run commands on that user / directory

There had been some progress on how these scans and processes work, but I'd still.love a dedicated way to force run a scan on a subset of the database, and having user and folder available as switches would be great.