AI Faces - How to reduce very large faces_tag.log?

9 views
Skip to first unread message

kyo...@dharma.net

unread,
Jul 1, 2026, 2:05:09 PM (3 days ago) Jul 1
to ResourceSpace
Hi - I'm having a very good experience overall with our new RS 10.7 on Ubuntu 24.04 and the AI Faces - Insightface plugin.

We have 27,360 Photos in our archive, processed as 72,256 total faces detected with 0 Images to process.

The plugin is very accurate, so we have set our Faces match threshold and Faces tag threshold both down to 54% with good results. I left the Face confidence threshold at the recommended 70% So far, I feel I am advised not to go much lower or risk getting errors in the matching.

I do have an issue where the faces_tag.log created by the cron job running faces_tag.php in /etc/cron.d/resourcespace is very large. I have 109,247 entries, most of them in the following format:

Processing face 3 in resource 37,584
No matching faces found

I have a hand-checked a number of these, and they do have a name in the Suggested Persons list I designated in AI Faces Configuration. 

Do I still need to run some kind of update on the database I am not yet familiar with? And how many of these log entries are created by the need to go through Photos and tag missing names manually? 

Is there a way to set up any kind of automated way of processing this list, or is this just the work of the Photo Archivist now?

Best - Frank
Am I getting 

kyo...@dharma.net

unread,
Jul 1, 2026, 2:09:13 PM (3 days ago) Jul 1
to ResourceSpace
And what do you think of the AI Search recommendation to "reduce my clutter" by combining all unmatched images as follows:

Remedy B: Give Unmatched Faces a "Trash" Node ID (The Permanent Fix)
The most elegant database strategy to clear the backlog permanently is to create a special placeholder person named "Unknown Face" or "Background Clutter" inside ResourceSpace, and mass-assign all unmatchable vectors to it.
  1. Go to your ResourceSpace UI under Admin > Manage metadata fields > Suggested Persons > Manage options.
  2. Add a new option named precisely: Unknown Noise.
  3. Run a quick SQL query to find the ref ID number assigned to that new option:
    sql
    SELECT ref FROM node WHERE name = 'Unknown Noise';
    Use code with caution.
    (Let's pretend it returns Node 9999).
  4. Now, run this single SQL update query. This will take all 109,247 empty face vectors and assign them to the trash profile:
    sql
    UPDATE resource_face SET node = 9999 WHERE node IS NULL OR node = 0;
Reply all
Reply to author
Forward
0 new messages