Check it out if you do art: https://glaze.cs.uchicago.edu/

  • TraschcanOfIdeology [they/them, comrade/them]
    ·
    2 years ago

    But ai can be trained to look for an obvious watermark. This seems less easy to identify, or at least for a person to tell the computer which artwork hasn't been glazed.

    • KnilAdlez [none/use name]
      ·
      2 years ago

      A deep learning model can look for a watermark, but is it worth it to remove the water mark or just leave it out of the training set? It's not a perfect solution, and if someone was really determined they could get around it, but it will be more trouble than it's worth a lot of the time.