An Abnormal Expression Detection System (AEDS) Using Deep Learning Algorithms

Abstract

In the last decade, many deep learning algorithms have achieved remarkable success and gained popularity in various computer vision tasks, including object detection, image recognition, and segmentation. This AEDS (Abnormal Expression Detection System)leverages the power of deep learning algorithms to detect abnormal facial expressions in real-time automatically. AEDS proposed two important models; those are Deep CNN and RNN. CNN is responsible for learning discriminative features from facial images and capturing local and global information. These features are fed into the RNN, which can effectively model the temporal dependencies present in facial expressions. To train the AEDS, a large dataset of facial expressions, comprising both normal and abnormal instances, is collected and labeled. Data augmentation techniques enhance the model's generalization capabilities and reduce overfitting. The deep learning model is trained using this dataset supervised, where the ground truth labels indicate normal or abnormal expressions. The evaluation of the AEDS is performed on an independent test set, and AEDS outcomes prove high accuracy and robustness in detecting strange facial expressions. Furthermore, the system exhibits real-time performance, making it fit for the many applications, for instance, security monitoring, healthcare, and human-computer interaction.