Facial Emotion Detection using CNN and OpenCV

In various cases, like social robotics, psychological research, and human-computer interaction, highly accurate real-time facial emotional recognition is highly desirable. Nonetheless, current methods suffer from a low degree of correctness, limited real-time, and sensitivity to pose or lighting conditions as well. Therefore OpenCV combined with Convolutional Neural Networks (CNN) can be used to offer a viable solution concerning these difficulties (Rahman et al., 2018). The system uses an already pre-trained VGGFace model to extract features from the detected face after preprocessing it to get the best CNN input in OpenCV's Haar cascade classifier. Henceforth, these features are presented through the Softmax classifier for differentiation into separate emotions. The proposed system has shown remarkable accuracy of 95.2% on the FER2013 benchmark dataset which demonstrates its ability to provide accurate emotion detection results. This software is very efficient because it can process data at a real-time rate of 25 fps on an average PC machine. Therefore, since this suggested system has a high accuracy alongside being ability to work in real-time, it may find significant application in various areas requiring emotion recognition (Taweh & Oyekanlu, 2017). Its high precision, real-time performance, and robustness to variations in lighting, pose, etc. make the suggested system a promising choice for different applications in social robotics, human-computer interaction, and other fields.