BEiT-3 is a foundation model that can be fine-tuned to detect fake images with a classification head.
Use https://colab.research.google.com/github/ with "Include private repos" to run fake_image_detection.ipynb
generate_balanced_dataset.py was used to generate a balanced train dataset: https://www.kaggle.com/datasets/dmitryvinnik/artifact-60k-balanced
The predictions are provided for https://www.kaggle.com/datasets/dmitryvinnik/rd-predict-dataset