Description
The goal of this project is to reduce the amount of bad logos (often violating copyright rules) on Commons by detecting them early in the upload flow.
Background: based on extensive analysis in T340546, logos represent the second top reason for deletion requests amongst copyrighted reasons.
The project contains two parts:
- Post upload detection of logos uploaded to Commons. Appropriate tagging to facilitate moderation of potentially bad logo. The detection is powered by the logo detection model: see the model exploration in T352748 and the project page
- [NTH - not in the MVP scope] Warning UI for uploaders during the Released rights step: See project page | designs
Use cases
As a moderator,
- I want to be able to easily access uploaded logos to Commons
to identify logos that might be violating copyright rules
As an uploader, when uploading a media file containing a logo through Upload Wizard on Commons,
- I want to place my file in the correct own vs not own work category
- I want to ensure that my file satisfies Commons rules and choose the right license option
So that it does not get flagged and deleted in the future.
Scope of MVP
The MVP scope contains:
- Serving the logo detection model on LiftWing
- Post upload job to detect logos uploaded to Commons T368624
- Tagging detected logos for easy moderation access (template TBD)
Outside of MVP scope
- UI to alert uploaders when logos are detected in Upload Wizard T363759
- Iteration of the logo detection algorithm
- Detection other types/classes of problematic images
- Interactive UI do display a warning
- Post-upload moderator UX