Question: Using this to detect traditional watermarks

[Serenity-Solstice]

Hello!

Do you think this code could be adjusted to help AI filter out artworks that are watermarked in more traditional ways like over-image watermarks and signatures? Additionally, do you think it could be used to identify and tag individual pieces with specific creators based on what the program reads? I think this code has great potential to get the AI art white knights to shut up on both sides of the spectrum if that's the case (those in-favor and against, I mean). While you're right that human artists take inspiration from other artists all the time, it's often acknowledged in ways that aren't nearly as prevalent, if possible at all, with AI art in its current state. Being able to recognize and credit artists from which the data was retrieved from would quiet a significant portion of the art community, since it would open the door to accurate reference credits. Happy to help if this might be a possibility!

[huzpsb]

I had the same problem and somehow managed to make a model
https://github.com/huzpsb/DeWm/