Real-time Facial Expression Recognition Research Based on Blazeface Face Detection and Resnet Emotion Classification

Facial expression recognition technology has the potential to enhance computers’ ability to comprehend and identify emotions, thereby improving human-computer interaction. In response to the real-time detection challenges encountered in current research, this paper developed a facial expression recognition system based on deep learning. Firstly, OpenCV is employed for preprocessing the acquired facial images, and the Blazeface face detector is utilized to provide the coordinate information of the faces within the images. Secondly, the paper utilizes the Resnet18 neural network to train the Fer2013 facial expression dataset for the purpose of developing a facial emotion classifier. Finally, the developed facial emotion classifier was utilized for the recognition of seven fundamental emotional states (anger, disgust, fear, happiness, neutrality, sadness, surprise), and the performance of the classifier was assessed and validated using publicly available expression databases. The experimental findings suggest that Blazeface face detection and Resnet emotion classification are appropriate for real-time detection of facial expressions.