DeepFake Detection Using Inception-ResNet-v2

Dataset

Dataset link: Kaggle

Preprocessing

• OPEN-CV was used to get frames from the video.
• Face Detector from Dlib was used to get face images from frames.

Model Accuracy

Inception-ResNet-v2 pretrained model was used to classify real or fake.

Model	Accuracy (%)
Inception-ResNet-v2	91.2%