Initial `classify` handles already tagged images

Which version of recognize are you using?

10.0.7

Enabled Modes

Object recognition, Face recognition, Video recognition

TensorFlow mode

Normal mode

Downstream App

Memories App

Which Nextcloud version do you have installed?

32.0.6 Turnkey

Which Operating system do you have installed?

TurnKey GNU/Linux 18.1 (Debian 12/Bookworm)

Which database are you running Nextcloud on?

10.11.14-MariaDB-0+deb12u2

Which Docker container are you using to run Nextcloud? (if applicable)

Proxmox LXC container

How much RAM does your server have?

5GB

What processor Architecture does your CPU have?

x86_64

Describe the Bug

Hello. I have an old Nextcloud setup with about 500G of photos.
It tooks more than a day to recognize them, so recognize:classify process dies probably because of PHP process timeout. During that day, it classifies about 35K of photos.
After relaunching a classification process, Recognize handles already classified from a first run photos ignoring a --retry key, so like 2/3 of files stay untouched. Tags count in the database stays at 35K without any changes despite the classify progress reports

Expected Behavior

Recognize skips classified photos and handles fresh ones

To Reproduce

Env quirks which may affect smth:

• Proxmox + Turnkey Nextcloud container image
• NC data folder is a symlink to a mounted secondary storage: /var/www/nextcloud-data -> /nextcloud-data

1. Run turnkey-occ recognize:classify
2. wait a process to die
3. rerun a process: turnkey-occ recognize:classify --retry

Debug log

No response

Start at the recognize:classify command and trace how the --retry option handles images that already have recognition tags. Reproduce the interrupted run and rerun with --retry in a comparable Nextcloud setup, then verify that previously classified photos are skipped while remaining photos receive tags and progress matches the database.