Speculative Analysis of CNN's and ResNet50 for the Identification of Emotion From Facial Expressions

Facial recognition software has become essential in several domains, such as security and surveillance and smart device user identification. This paper looks at the effectiveness of deep learning methods for facial recognition to classify images based on the emotion, with a particular focus on Convolutional Neural Networks (CNN). A condensed CNN and ResNet50 are the two CNN architectures whose impacts on loss values, processing efficiency, and accuracy are specifically evaluated. A variety of face datasets containing images depicting seven distinct emotional states—Angry, Disgust, Fear, Happy, Sad, Surprise, and Neutral—are used to train and assess the models. Assessment metrics, such as calculation time, accuracy, and loss values, are employed to evaluate the effectiveness of every technique.