RECOGNITION OF HUMAN FACES IN VIDEO CONFERENCE APPLICATIONS USING THE CNN PIPELINE

Abstract

The COVID-19 pandemic has forced daily face-to-face activities to be carried out online using video conferencing applications. To record participant participation in meetings using a video conference application, an online form application is used. However, participants sometimes do not see this and are often missed due to the large number of incoming chats. Therefore, the use of face detection for attendance using a combination of CNN to detect all the faces in a video conference using YOLO Face and CNN to recognize the owner of a face using Smaller VGG in a pipeline will make it easier to recognize participants who are present at the video conference. The results of the Smaller VGG training are obtained, namely the loss value of 0.059, the accuracy value is 0.995, the recall value is 0.994, the precision value is 0.996. Meanwhile, for the validation phase of the model, the loss value is 0.497, the accuracy value is 0.979, the recall value is 0.979 and the precision value is 0.981. In terms of training duration, the smaller VGG has a duration of 4 minutes and 16 seconds. The Smaller VGG model was combined with YOLO to create a CNN pipeline and was successful in recognizing the faces of video conference participants