本研究使用深度學習來訓練一個能夠判斷人像影片真偽的神經網路模型—輸入一支帶有人像的影片，該模型會標示此影片的人像是否經過偽造。本研究以FaceForensics++ 為基礎，擴充其神經網路模型的功能，並且使用亞洲人像當作神經網路模型的訓練資料，使神經網路能在辨識亞洲面孔時有較優秀的辨識能力，同時也針對高、中、低解析度的圖片訓練，使其面對不同像素的圖片、影片仍存在良好的辨識能力。

This study designs a neural network model by deep learning, which integrated the knowledge in FaceForensics++ and their experimental results. Given a video containing people, this model is capable of determining whether the people in the video have been deepfaked. In the proposed model, we used FaceForensics++ as the design basis and extended the functionalities to enhance the ability of recognizing Asian faces from the deepfake videos. We also trained the model on high, middle and low-quality images to ensure that it is able to recognize images and videos with different qualities.