Self-Cure Networks with Latent Distribution Mining for Facial Expression Recognition

Facial expression recognition is designed to identify human emotions from facial features. Relevant investigations show that deep learning has higher universality than manual design features and shallow learning methods. Unlike a unified laboratory environment, an unrestricted real-world environment often contains distractions such as lighting resolution, occlusion changes in head posture, and individual character differences, making it more challenging to annotate large facial expressions. To solve the above problems, this paper proposes a Self-Cure Networks with Latent Distribution Mining(SCN-LDM) based on latent distribution mining based on deep learning technology in order to fully mine the image and label information of facial expression data, suppress the label ambiquity of large-scale data sets, and acquire robust facial expression features with expression discrimination ability.