Recognize index via occ just for NEW photos

🍱 Features & apps
1

Hey all!

I’m still new to NC and all, but I have everything running nicely. One issue where I’m stuck (well, extremely slowed down) is Recognize.

I’ve added 100.000 photos, and ran the initial face scan with occ. But that ran several hours then threw an error. I’ve re-ran the scan few times, but it always starts from beginning, and eventually it dies (not necessarily with same error or in same place). In the meantime cron jobs indexed a lot of faces as well, so my collection is filling up nicely. But waiting for cron and batches will take weeks or months.

My question is - can I force a scan using OCC tools on photos that have NOT been scanned at all.

Eg. most other scans (file scan, memories index, preview generator) know what they’ve scanned so far, and when you manually re-run them they will fly through the list of files, and only slow down to index/scan/generate info on files they encounter for the first time. So for example file scan can take 2 days to scan everything, but once done, manually re-scanning takes under 10 minutes.

On the other hand, running “recognize:classify” always starts at the top of file list, and always takes same amount of time to go through pics. Other apps have flags like “-f” (force), but seems that for recognize that’s default behavior.

Any way around the limitations? Am I missing some other command or a function?

I just want it to work and work until done, and if it dies (or I need to restart server or SSH session) that I can point it to next folder or something … I’d be so happy to get this going somewhat faster… :slight_smile:

2

Nevermind, no one knows or there’s no workaround. Luckily, last night a run of classify went through without errors (probably because finally there were no other large indexing jobs running in parallel). So my initial scan is finished and I won’t need the workaround in the foreseeable future (hopefully)… will just hope that next large batches doesn’t fail during night runs.

Still, if someone stumbles on this question and has any recommendations or ways to run classify on a single folder or something similar, please let me know. I’m sure I’m not the only one doing occasional large imports of photos.

Example workflow (using CLI/OCC):

  • copy 10.000 photos to “uploads-new” folder (ideally using direct copy in OS, for speed)
  • run files:scan --path=“/user/uploads-new”
  • run files:scan --all --generate-metadata
  • run memories:index
  • run maps:scan-photos
  • run preview:generate-all --path=“/user/uploads-new”
  • run recognize:classify ( --path=“uploads-new” )
  • run recognize:cluster-faces
  • move 10.000 photos to “Photos”, leave “uploads-new” empty for next batch
  • done (and you have new photos in Files, Photos, Memories, Persons, Maps, etc, all indexed and ready)

Example workflow would be done in a few hours, and works pretty good for me so far. I just don’t see how to run classify with something like “–path” to speed up my face classification :-/

1 Like