Facial expression recognition is an interesting research area that facilitates computers to apprehend human emotions for quantitative analysis of human emotions. Facial expression recognition is a significant research concept in pattern recognition, computer vision, and artificial intelligence both at home and abroad.

Facial expression recognition owns wide applications in diverse domains such as human-computer interaction, medical treatment, self-driving, virtual reality, augmented reality, advanced driver assistance systems, education, and entertainment. Automated facial expression recognition is a challenging problem in computer vision.

For machine learning algorithms, recognizing facial expressions is a complex problem. With the outstanding success of deep learning, several types of architectures are exploited for automatic facial emotion recognition to perform better. In the field of computer vision, deep learning methods have remarkable efficiency. The deep learning architectures for automated facial expression recognition impart fine-tuning of real-time facial data with a higher accuracy rate. Deep learning-based automatic facial expression recognition obtains better detection and can test and train on static or dynamic sequences images. Some of the impressive deep-learning models for facial expression recognition are;

•  Convolutional Neural Network (CNN): CNN is highly employed for facial expression recognition. CNN is potent to face location alterations and scale changes and better for identifying unseen face pose variations. CNN effectively contends the issues of translation, rotation, and scale invariance in facial expressions recognition. Region-based CNN, faster R-CNN, 3D CNN, and CNN with Attention mechanism recently emerge CNN derived models for facial expression recognition.

•  Deep Belief Network (DBN): DBM models are combined with various features and classification models for better recognition performance.

•  Long Short-Term Memory (LSTM): LSTM models help appraise dimensional representations of emotions in audio and visual applications in facial expression recognition. Recently, 3D Inception-ResNet (3DIR) has been combined with LSTM to recognize video sequences in facial expression recognition.

•  Generative Adversarial Network (GAN): GAN models are applied for pose-invariant and identity-invariant expression recognition in facial expression recognition. Recently developed GAN models for facial expression recognition are multi-task GAN, Representation-Learning Variational Generative Adversarial Networks, conditional Generative Adversarial Networks, and end-to-end GAN-based models.