Emotional Nuances: A Comparative Study of Convolutional Neural Network and Residual Network

Despite having a critical role in human-computer interaction and desirable computing, emotion detection has not reached its full potential. Our research presents a careful comparison of two of the most outstanding deep learning networks, CNN and ResNet, in the field of emotion recognition. For better understanding, they are trained on a specific dataset for emotion detection. Where ResNet has a very slight advantage in terms of accuracy, it is excellent to depict subtleties of emotion while CNN has strong basic computational ability. In addition to discussing the accuracy of CNN and ResNet and using them in practice, this study examines their calculation speed, which forms a primary criterion even in cases with limited computing resources. Real-time issues succeed in underlining the conflict between precision and efficiency, which is also beneficial for application in constantly changing conditions. With current and future developments of the text emotion detection, this research adding to the pool of data or information to enable the researcher and practitioner to make wise decisions regarding the selection of proper architecture for given application.