Deep Fake is a kind of fake image or video created using artificial intelligence to superimpose thefaces of the targeted person to any other image orvideo with extreme precision that seems very original and impossible or very difficult to detect with normal human eyes. Basically, it is developed with the help of deeplearning using AI to perceive the original facial recognition of the targeted person and matching the same while putting on the fake images or videos. Apart from politicians, to create a piece of sensational fake news among the audience, Deepfake is also creating content in the adult industry to target popular celebrities. They attack mainly female personalities to gain the attention of online audiences to watch such images or videos and share with others. Though, few people can do it as a revengeful act to defame the popular personalities from the entertainment industry. Mainly in erotic videos, the face of the targeted celebrity is superimposed using the AI technology that becomes very original among the audience.
This project aims to detect deepfake videos by preprocessing face-only videos using the face_recognition library and implementing a deep learning model with the Resnext-50 architecture. Leveraging advanced computer vision techniques, the model is fine-tuned for high accuracy in identifying manipulated media, showcasing its potential for real-world applications in detecting fake content.