Relative Patch Response Learning for Generalizable AI-Generated Image Detection

Key points

• Existing detectors are typically trained on either independently collected real and generated images or aligned real-generated pairs designed to mitigate content bias.1
• Building on aligned pairs, recent methods form a mixed view by replacing some patches of the real image with their generated counterparts.1
• However, we find that self-attention lets real and generated patches interact, so the feature of each patch no longer reflects its own source alone.1
• To this end, we propose Relative Patch Response Learning (PRL).1

Sources (1)

• [1]Relative Patch Response Learning for Generalizable AI-Generated Image Detection
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 12:49 PM

  Generative models can now synthesize highly realistic images, simultaneously increasing the risks of misinformation and visual forgery.

  Existing detectors are typically trained on either independently collected real and generated images or aligned real-generated pairs designed to mitigate content bias.

Extractive summary: sentences quoted from the sources.