Create a post-upload job to rescale uploaded images and test them against the LiftWing logo detection model
If they're logos then use wbsetclaim to add instance of:logo with the qualifier determination method:machine learning
We'll probably also need to check WikibaseMediaInfo to make sure instance of gets written to the search index
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T349641 [EPIC] MVP Logo machine detection on Commons | |||
| Resolved | mfossati | T368624 [XL] Post-upload job to detect logos |
Note that this should NOT be part of UploadWizard, because we want the detection to happen for all uploads. Note that non-WMF installs won't have access to the logo detection endpoint
So we think we'll proceed like this for now (simplest approach, because it's a PoC*)
*for a non-PoC version we think we probably ought to have a non-MW api endpoint (written in node?) where you can send it an image/thumbnail and get back a likelihood-of-deletion score (plus other information)
References:
Blocked by T370861: [L] Refine the logo detection model: a monitoring job is currently running on Toolforge via https://toolsadmin.wikimedia.org/tools/id/gogologo, but we'll need to switch to the refined model.
Plugged secondary models from T370861: [L] Refine the logo detection model into the monitoring job.
Closing, follow-up expected after T370137: Set up a bot account on Commons to add logo-detection statements 's resolution.