A Detect method for deepfake video based on full face recognition
Kai Ping Feng, Jan Wu, Min Tian · 2020
In recent years, with the continuous upgrading of computer hardware and the continuous development of deep learning technology, new multimedia tampering tools can make it easier for people to tamper with faces in videos. Tampered videos produced by these new tools may hardly be detected by human, so we need effective method to detect these face-tampered videos. Current popular video face tampering technologies mainly include Deepfake technology based on self-encoder and Face2face technology based on computer graphics. In this paper, we propose a new method for tamper video detection based on the full faces. Facenet algorithm is used here to compare the similarity between real and fake video faces. Finally, in the experimental part, the results showed a significant effect.