Facial Expression Recognition Using Deep Learning Model on Crowd

In the field of computer vision, Facial Expression Recognition (FER) is an essential problem that automates emotion analyzing and understanding in human facial expressions. Facial occlusions, changing lighting conditions and overlapping facial attributes in crowded environments make it difficult to recognize the expression of emotions. In this research we are solving a problem of group friendly facial expression recognition using deep neural network models. We also tested a number of complex deep neural network architectures to learn fine-grained patterns and representations in human face videos or images. In training and testing these models, resources such as AffectNet which encompasses a broad spectrum of people with various facial expressions is utilized to make them robust against most general cases in crowds. The results of the deep neural network model are presented, revealing its ability to detect facial expressions efficiently with high accuracy and recall. This is supported by the performance of their models being around 80–94 % in terms of accuracy and F1-score, based on different studies. Their benefits involve deploying state-of-the-art architectures and performing robustly on large scale datasets; fusing multi-modal for improved recognition, learning temporal relationships will be well established technologies; scalability to resource-constrained environments.