<p> Emotion recognition from human facial images is one of the important <br /> directions for human-computer interaction, security systems, mental health <br /> monitoring, and intelligent systems. Especially in the development of humanoid <br /> robots in the field of human-computer interaction, one of the important features <br /> of a humanoid robot is that the robot can communicate with humans while sensing <br /> their emotions. The development of emotion recognition systems is a significant <br /> challenge in computer vision and deep learning. In this study, we suggest a good <br /> way to recognize emotions from people's facial images by using well-known <br /> convolutional neural networks that have been trained on a better version of <br /> the FER-2013 dataset. In this study, the effectiveness of implementing the task <br /> of emotional state recognition using advanced convolutional neural networks <br /> is studied. In particular, the results of the popular pre-trained architectures <br /> ResNet-50, VGGNet-16, DenseNet-121, and EfficientNet-B0 on facial expression <br /> recognition are analyzed. The study used an improved and augmented version <br /> of the FER-2013 dataset. During data processing, imbalances between facial <br /> expressions, low-quality images, and incorrectly labeled images were detected <br /> and corrected. In addition, data augmentation techniques were used to reduce the <br /> problem of overfitting. The model results were evaluated based on criteria such as <br /> accuracy and loss function.</p>