Analysing Facial Dynamics - Navarasas in Kathakali Videos Using Deep Learning

Indian classical and folk dances are visual treats for the audience; the expressions used by the artist are crucial in conveying the narrative of the accompanying music to the audience. Kathakali – a classical dance rooted in the southern Indian state of Kerala, combines storytelling with eye-catching attire, face masks, and makeup. Facial expressions and hand gestures play an essential role in this process, as they are the primary means of communication between the dancer and their audience. Kathakali is widely prevalent among Kerala attractions for tourists, entertainment, and temple rituals. Fathoming the facial expressions of Kathakali artists is crucial for fully conceiving and appreciating the art form. Facial expressions in Indian Classical dance forms called Navarasa use a range of facial expressions through which a dancer can connect with their audience and draw them into the story they are trying to tell. The process of identifying and understanding Kathakali expressions has been automated to facilitate recognition and analysis. In addition to aiding in the preservation of cultural heritage, the focus has been placed on identifying and classifying the facial expressions used in Kathakali performances. A combination of ViT and LSTM networks was applied to observe Kathakali Bhava, or facial expressions, from videos, achieving an accuracy of 96.67%.