MinaAlmasi
commented
4 months ago Hi!
The stable diffusion code is not available as the dataset was not created by me. The README references the paper by Bird and Lofti (2023), but I might need to make that even clearer - apologies for that! The code was created for an exam, so I have not maintained it as a normal repository! Anyways, you can find the paper and dataset here:
J. J. Bird and A. Lotfi, "CIFAKE: Image Classification and Explainable Identification of AI-Generated Synthetic Images," in IEEE Access, vol. 12, pp. 15642-15650, 2024, doi: 10.1109/ACCESS.2024.3356122.
Link to paper: https://ieeexplore.ieee.org/abstract/document/10409290 Link to dataset: https://www.kaggle.com/datasets/birdy654/cifake-real-and-ai-generated-synthetic-images
Regarding the discrepancy in Figure 2 compared to CIFAR-10, I assume they wanted to present the images in high quality for visual appeal. They mention in the paper that they render the images in high resolution but downsize them to match CIFAR-10:
"Synthetic images were rendered at a resolution of 512px before resizing to 32px by bilinear interpolation to match the resolution of CIFAR-10."
Hope this cleared up the confusion. Don't hesitate to reach out again if not!
Hello,
Could you please clarify where the stable diffusion part is located? Additionally, I noticed that Figure 2 is high resolution, while CIFAR-10 is typically low resolution. Could you explain the reason for this discrepancy?
Thank you!