Context

The motivation behind the creation of this dataset is to have a challenging Test set for the task of classifying fake and real human faces. Most of the available datasets on Kaggle are "Uniform" and doesn't present a good variance of face features, particularly for the "Fake" class. Moreover, the fake faces collected in this dataset are generated using the StyleGAN2, which present a harder challenge to classify them correctly even for the human eye. The real human faces in this dataset are gathered so that we have a fair representation of different features(Age, Sex, Makeup, Ethnicity, etc...) that may be encountered in a production setup.

Content

The images available in this dataset are in a JPEG format and of uniform size of 300x300. There "Fake" faces are collected from the website thispersondoesnotexist.com. There "Real" faces images are collected through the API of the website Unsplash and then the faces are cropped out of using OpenCV library.

Total number of images: 1288 Number of "Fake" faces: 700 Number of "Real" faces: 589

The data.csv contains the images Id and the corresponding label.

Inspiration

Can you achieve a high accuracy on this dataset?