Visualizing Sentiments: Facial Expression Recognition in the World of Kathak Dance Using Deep Learning

This research focuses on facial expression recognition in the context of Kathak dance, a traditional Indian art form known for its emotional storytelling. A novel methodology, “ExpressNet,” combines transfer learning and advanced machine learning techniques to classify the facial expressions of Kathak dancers. Using a diverse dataset of Kathak dancers affiliated with Ntrutyankur Kathak Dance Academy, the study enhances emotion detection resources for this art form. The approach involves preprocessing, hierarchical feature extraction, and input into classification models: support vector machine, random forest, and a fine-tuned, fully connected network. Principal component analysis aids SVM and random forest to visualize decision boundaries effectively. “ExpressNet” achieves a 69.44% accuracy in discerning Kathak facial expressions, contributing to emotion analysis in performing arts. This advancement has implications for human–computer interaction, virtual reality, and emotion-influenced interactive systems, offering a valuable asset for Kathak facial expression recognition studies.