In this project we used Deep Learning to detect fake images generated by FaceApp app gender swap feature.
FACEAPP is a famous app that uses Deep Learning to generate fake pictures of people with different styles of hair, makeup and so on. But in this project we are interested in its gender swap feature which as its name indicates turns images of males into females and vice versa.
The goal is to scrap a decent amount of fake and original images, clean the images and train a deep learning model using transfer learning and FASTAI to achieve a good accuracy.
dataso that we can clean it later.
more_originaland were finally added to the
originalfolder before training.
oon the keyboard it means the selected area contains real person and it's moved to the original folder
fit means the selected area contains a fake image generated by faceapp and moved to the fake folder
qit will take us to the next image
I used FASTAI to train my models
I tried a resnet34 and resnet50
I found that resnet34 produced better results So,
Don't forget to tag @moaaztaha in your comment, otherwise they may not be notified.