<![CDATA[Face detection is a main and important process in the field of face recognition that has been widely studied. The purpose of face detection is to determine the presence and mark the position of faces, in both images and videos, called bounding boxes. One important problem in face detection is to differentiate between face spoof and non-spoof which is referred to as face spoofing detection. Face spoofing detection is an important task used to ensure the security of face-based authentication and facial analysis systems. Therefore, we need a model that can detect face spoofing. In this paper, the process to build a model that can be used to detect face spoofing on video is carried out using Faster R-CNN with Resnet50 architecture. Faster R-CNN is one of the superior algorithms in solving various object detection problems. The dataset used in this paper is a Replay-Attack Database, provided by Idiap Dataset Distribution Portal.The training phase used 360 videos, consisting of 300 spoof videos and 60 non-spoof videos. The average accuracy of the training stage is 97,07% with a total of 21 epochs. The test results show that the resulting model successfully determined bounding boxes and detected face spoof and non-spoof on the video effectively.]]>
Copyrights © 2020
