Analyzing Facial Emotion Patterns in AffectNet with Deep Neural Networks

We have developed an emotion detection system using convolutional neural networks (CNNs) trained on the AffectNet dataset. Emotion recognition is essential in human-computer interaction and affective computing. The AffectNet dataset offers diverse annotated facial expressions for effective model training. Our approach involves constructing three distinct CNN architectures with varying convolutional and pooling layers to optimize feature extraction and emotion classification.To evaluate performance of our models, we used accuracy, precision, recall, auc-roc score and a confusion matrix, providing insights into the effectiveness of each CNN architecture in classifying facial expressions according to different emotions. This approach showcases the potential impact of CNN-based methods in enhancing user experiences and interaction in real-world applications. Further refinement and analysis of these CNN architectures can lead to improved emotion detection capabilities and broader applications in various domains.