Currently working on a research paper where I’m basically teaching a model to answer:
“Is this scientific image real, AI-generated, or secretly tampered with?”
Using a CNN + Vision Transformer hybrid.
Paper still in progress.
GPU already regretting it. 💀