<![CDATA[Recognizing facial expressions is a critical aspect of computer vision and human-computer interaction. It facilitates the interpretation of human emotions from facial images, aiding in applications such as affective computing, social robotics, and psychological research. In this work, we propose using hybrid deep learning models, ResNet50 and ResNet34, for facial expression classification. These models, pre-trained on large-scale datasets, demonstrate exceptional feature extraction capabilities and have achieved excellent performance in various computer vision tasks. Our approach begins with the collection and preprocessing of a labeled facial expression dataset. The collected data undergoes face detection, alignment, and normalization to ensure consistency and reduce noise. After preprocessing, the dataset is divided into training, validation, and testing sets. We fine-tune the ResNet50 and ResNet34 models on the training set, employing transfer learning to adapt the pre-trained models specifically for the facial expression recognition task. Optimization techniques such as SGDM, ADAM, and RMSprop are used to update the models' parameters and minimize the categorical cross-entropy loss function. The trained models are evaluated on the validation set, achieving an accuracy of 98.19%. Subsequently, the models are tested on unseen facial images to assess their generalization capabilities. This proposed approach aims to deliver accurate and robust facial expression classification, thereby advancing emotion analysis and human-computer interaction systems.]]>
