REAL-TIME FACIAL EXPRESSION RECOGNITION BASED MACHINE LEARNING WITH WEBCAM

• Main Author: Ferdi Ghozali, Mokhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39973
• Institution: Institut Teknologi Bandung

Facial expression is one of many ways for humans to interact with each other. Thus, emotions are essential in face to face interaction. Robots will be able to interact better with human if they are able to recognize emotions. Artificial intelligence can be modelled to enable robots to recognize human emotions. Facial expression recognition start with taking image from webcam, than detecting face in te picture and then classify into basic emotion class. The related research on face detection was used Histogram of Oriented Gradient (HOG) with Support Vector Machine (SVM) and Convolutional Neural Network (CNN) method. The related research on facial expression recognition was used CNN and facial landmark with SVM. So far, studies regarding face emotion expression recognition has been done on face images in controlled environments. Affectnet is a dataset containing facial expression images in an wild environment, and is the largest of its kind. The method proposed in this paper is fine-tuning in pre-trained VGG architecture with imagenet as the initial weight. This method managed to achieve accuracy of 56%. This method provides a fairly rapid increase where the accuracy of the previous method was only 44.05%. This study also tried to do threading so the system can run in real time.