Optimized Deep Learning based on Facial Emotion Recognition to Analyze the Efficacy of Online Classes

Formal classes have allowed students to transcend barriers brought about by distance and time, and have brought in students from various geographical places and backgrounds seeking a great education. Nevertheless, monitoring students' participation and emotional well-being has posed fresh challenges in online classes. The present study will analyze students' facial expressions during online classes, understand their emotional states, and assess a fine-tuned MobileNet V2 model. The designed technique used the CK + dataset of labeled facial expression data collected in a controlled laboratory. Pre-trained on ImageNet, a big-scale image classification dataset, then fine-tuned on the CK + dataset for the specific emotions when learning online classes, the model MobileNet V2 should recognize these emotions expressed by students through that process. In the end, the parameters for the MobileNet V2 technique are optimized using the updated Lichtenberg Optimization Algorithm (LOA) in the DL model. Also, improve the generalization capability of the model using preprocessing techniques like image augmentation and normalization. The initial fine-tuned performance level achieved from a pre-trained model was average; after fine-tuning, the performance increased to an accuracy level of 98.90%, indicating enhanced detection and classification accuracy of facial emotions in online classes.