The Constrution And The Application Research Of The Facial Expression Datasets In The Wild

Facial expression recognition is a hot topic in the artificial intelligence and human-computer interaction fields. The goal of facial expression recognition is to make the computer automatically recognize peoples' expression, then, analyze their emotions, so that the computer can serve people better. Facial expression reflects and transmits peoples' thoughts and feelings, contains rich and subtle emotional change, analyze and study it helps to some application fields such as human-computer interaction and affective robot.Generally speaking, there are two tasks in face recognition: face identification and face verification. However facial expression recognition is another important topic in this filed, challenge happens in face recognition exists in facial expression recognition, and maybe more difficult to resolve. Here in my paper did some work on facial expression recognition, the related work include the following aspects:(1)Facial expression recognition got lots of researchers' attention, but a big partial focus on facial expression recognition methods, and they get less focus on the facial expression database. Specific to the problems existed in the current public face databases that used to facial expression recognition, this paper construct a facial expression database. The new database includes various kinds of facial images in the wilds, such as different age periods(7 periods), different races(4 races) and different skin color(4 kinds) and so on. We also labeled the new database with two kinds of labels, 10 kinds of emotion labels indicate the facial expression and five partial labels indicate the status of the five components of the face.(2) Use traditional algorithm to verify the new database. Use the two kinds of labels to do some comparison experiments, and the results show that these two kinds of labels help to facial expression recognition.(3) Designed and realized 3 CNNs for facial expression recognition. After the model realized, we use the new database to train the model. We use these three models to do some experiments and compared the results with the new model. Discovered through comparative analysis the second model acts well on the facial expression recognition task.