Abstract
The imbalanced number and the high similarity of samples in expression database can lead to overfitting in facial recognition neural networks. To address this problem, based on edge computing, a facial expression recognition method using deep convolutional neural networks is proposed. In order to overcome the shortcoming that circular consensus adversarial network model can only be mapped one-to-one, we construct a constrained circular consensus generative adversarial network by adding class constraint information. Discriminators and classifiers in this network can share network parameters. In addition, for the problems of unstable training and easy to encounter model collapse in original GAN networks, this paper introduces gradient penalty rule into discriminator’s loss function to achieve the normative constraint on gradient changes. Using this network not only generates sample data for a few classes in the training set of expression database, but also performs effective expression classification. Compared with other methods, the improved discriminative classifier network structure can enhance the diversity of samples and get a higher expression recognition rate. Even if other expression feature extraction methods are used, the higher recognition rate can still be obtained after using proposed data augmentation framework.
Highlights
With the rapid development of intelligent information technology and the wide use of computers, people can turn the complicated work to computers, which changes the traditional way of life, and provides great convenience for human beings
2) Aiming at the problems of unstable training and easy to encounter model collapse in original GAN networks, this paper introduces gradient penalty rules into the loss function of discriminator, which can achieve the normative constraint on gradient changes
We introduce gradient penalty rules into the discriminator’s loss function which can regulate the gradient change
Summary
With the rapid development of intelligent information technology and the wide use of computers, people can turn the complicated work to computers, which changes the traditional way of life, and provides great convenience for human beings. Different from using simple geometric transformations and crop for data augmentation, GANs introduce an adversarial loss function and learns the facial expression images with the same distribution as target datasets, which can solve the high similarity problem of generated samples. When there is a one-to-many mapping relationship (such as a neutral expression to a variety of expressions such as happy, sad and surprised), the model needs to be trained multiple times, which brings a huge time cost To address this problem, this paper improves Cycle GAN and further proposes a constrained circular consensus generative adversarial network for facial expression recognition. All of these mechanisms have achieved good recognition results, but the quality of images generated by GANs network is uneven due to the lack of constraints It can’t realize flexible expression mapping in the case of unbalanced number of facial expression samples. We can first use unlabeled samples to train GANs, and use a small part of labeled data to train discriminators for traditional classification and regression tasks
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
