MultiClass_DeepFake

GAN, Stable Diffusion, Stable&GAN Manipulation, Real

1. Real_Fake face identification refers to the detection of fake or synthetic faces; it usually involves identifying artificially generated faces or manipulated images that may be used to deceive facial recognition systems.

2. Detecting fake or synthetic faces can be challenging because modern techniques have become quite sophisticated. However, researchers are continuously developing algorithms and methods to tackle this issue. These methods often involve analyzing facial features, texture inconsistencies or using machine learning approaches to differentiate between real and fake faces.

3. It's worth noting that the technology and techniques in this field are continually evolving, and it's advisable to consult the latest research and advancements in the field for the most up-to-date information on real-time face identification and the detection of fake faces.

4. I have prepared the dataset that consists of 40K images. In total, there are 4 classes (Real, GAN, Stable Diffusion, GAN&Stable Manipulation ).

5. Real 10K Samples retrieved from the link https://www.kaggle.com/datasets/xhlulu/140k-real-and-fake-faces.

6. GAN 10K Images downloaded from the https://thispersondoesnotexist.com/.

7. Stable 10K images generated with different prompts like "A pretty face of a attractive man with style of the man".

8. Stable&GAN Manipulation 10K images retrieved from https://www.kaggle.com/datasets/selfishgene/synthetic-faces-high-quality-sfhq-part-4.

9. I have achieved an accuracy of 99.90% for 4K Test images(Diffusion_Fake:997 GAN_Fake:1010 Real:997 Stable&GAN_Fake:996).