Comparison and implementation of algorithms for manipulating face images from video

U ovom radu bilo je potrebno kreirati deepfake video. Na početku je opisan softver koji će se koristiti, te algoritmi pomoću kojih se kreiraju deepfakeovi. Deepfake video se u cjelosti kreirao pomoću DeepFaceLab softvera. Prvo su pronađeni modeli koji će se koristiti za video, nakon čega su kreirane njihove datoteke lica. Te datoteke su korištene za treniranje deepfake modela, koji se kasnije spaja sa slikama modela čije lice se mijenja. Konačan rezultat je spajanje svih slika u deepfake video.In this paper, the goal was to create a deepfake video. At the beginning, the software that will be used and the algorithms used to create deepfakes are described. The deepfake video was created entirely using DeepFaceLab software. First, the models to be used for the video were found, after which their face files were created. These files were used to train the deepfake model, which is later merged with the images of the model whose face is being changed. The final result is the merging of all images into a deepfake video