ANALYZING FACIAL EMOTION PATTERNS IN AFFECTNET WITH DEEP NEURAL NETWORKS

Abstract

The significance of emotion recognition and its potential benefits are explored in the research. Emotion recognition is crucial as it enhances our understanding of human affect and facilitates improved interactions between humans and machines. The study underscores the importance of accurately identifying emotions in various fields, including stress detection in person, user experience design, human-computer interaction, and social robotics. The research employs the AffectNet dataset, a large-scale repo of facial images equipped with emotion categories specifically to train and examine deep neural network models. Specifically, Convolutional Neural Networks (CNNs) are used due to their efficiency of effectively analysis tasks in image. The models are carefully designed as well as fine tuned to handle the variety and complex nature of the AffectNet dataset, Handling issues such as changing stances, clarity, and face limitations. The work provides a comparison of several CNN designs and their performance in emotion identification tasks. Each model's strengths and limitations are evaluated using evaluation criteria like as accuracy, precision, recall, and the F1-score. The results of this comparison work will help to improve the accuracy and efficiency of emotion identifying systems. The results of the research give useful information on the effectiveness of different CNN models in recognising facial emotions. Such results can help to develop emotion identification system, making them more useful in real-world scenarios. Furthermore, the work provides to the area of deep learning by identifying practical problems and solutions for training models on large and heterogeneous datasets such as AffectNet. The findings improve our knowledge of human emotions and establish the framework for future advances in personalised applications, therapeutic tools, and human-machine interactions.