Facial Expression Recognition Based On Deep Learnin

With the continuous development of computer technology and artificial intelligence technology,human-computer interaction technology has become one of the important research directions in the development of the information industry.Compared to the past,people today have higher requirements for human-computer interaction technology,requiring robots to be closer to people’s needs and provide appropriate feedback.To achieve this goal,robots need to have the ability to analyze emotions and provide more appropriate responses through emotional analysis in communication with people.Facial expression recognition technology is a key research field in emotion recognition technology,and is currently one of the hottest topics,with a wide range of application prospects.Due to the current lack of mobile devices for facial expression recognition during the rapid driving process of vehicles,this study conducts research from the following aspects based on the research of relevant literature at home and abroad:Firstly,the research background and significance of facial expression recognition technology,the research status at home and abroad,the basic principles of deep learning,the main learning framework,algorithms,representative models,and evaluation indicators of deep learning are comprehensively described.Secondly,this study imports the RAF-DB dataset into the Yolov 5 series of models for training,and compares and analyzes the results.Through analysis,it is found that although the training results of Yolov5 S are not as good as those of other models,due to its minimal model size,it can seek a balance between lightweight,fast reasoning,and accuracy,and is suitable for application scenarios with high requirements for real-time performance and resource constraints.Next,the YOLOv5 s model was lightweight using Ghostnet,Mobilentv3,and Shufflenet methods.The experimental results showed that the Ghost method was used to lightweight the YOLOv5 s model.By replacing the original convolutional layer,the lightweight model was reduced to 54.6% of the original model size.Finally,this study will reduce the weight of the yolov5s＿ghost model was quantified and deployed to the Linux platform for verification,proving the feasibility of lightweight,quantization,and deployment of facial expression recognition technology,making a certain contribution to the future application of facial expression recognition technology on mobile devices.