Machine learning classifies stances in Bharatanatyam dance. A team of researchers from Chennai's Anna University used state-of-the-art computational methods to accurately identify and classify 108 basic Bharatanatyam dance stances. AI technology can also model and preserve other traditional performing art forms.
This research explores the specialized field of human action recognition, with a focus on identifying poses in Indian classical dance, particularly Bharatanatyam. In dance, 'Karana' refers to the harmonious and melodic movements of the body, hands, and feet as described in the Natyashastra.
Karana is a combination of nritta hasta (hand movements), sthaana (body posture) and chaari (leg movements). The Natyashastra enumerates 108 karanas, illustrated in the elaborate stone carvings that cover the Nataraj temple at Chidambaram, revealing the link between these movements and Lord Shiva. In Bharatanatyam, it is difficult to automate posture detection due to the large number of different hand and body postures, mudras (hand gestures), facial expressions, and head movements.
This work uses automation and image processing techniques to reduce this complexity. The proposed approach has four steps. Skeletonization and data augmentation techniques for image acquisition and preprocessing, feature extraction from images, dance pose classification using a deep learning network-based convolutional neural network model (InceptionResNetV2), and point clouds for 3D model visualization. Creating a mesh from. .
Identification is simplified using cutting-edge technologies such as deep learning networks and the MediaPipe library for body keypoint recognition. A key phase, data augmentation, improves model accuracy by expanding small datasets. Efficient recognition of complex dance motions by convolutional neural network models facilitates analysis and interpretation. This creative method facilitates pose recognition in Bharatanatyam and establishes a standard for improving efficiency and accessibility for practitioners and researchers of Indian classical dance.
Given the diverse applications of human pose detection in daily life, significant difficulties have arisen in computer vision. Therefore, identifying postures in Indian classical dance, especially Bharatanatyam, is essential as it can affect human well-being.
The authors of this paper introduce InceptionResNetV2, a unique deep learning network-based convolutional neural network model. The model works based on the key aspects found in MediaPipe to accurately classify 108 dance positions. Their strategy was developed after a thorough analysis of relevant published literature.
Their design is based on extracting depth and spatial elements from photos separately and using both sets of information to recognize poses. Their architecture benefits from this special strategy, allowing it to better identify poses, as first proposed in their methodology and subsequently confirmed by comparison and result analysis in their work. I did.
Additionally, their feature extraction approach allows the proposed design to support a variety of positions. The main goal of future research projects is to improve performance through hyperparameter tuning.
Finally, their research has greatly aided the ongoing efforts to identify the stance of Indian classical dance, especially in Bharatanatyam. Their work improves the accuracy and robustness of posture recognition in this complex dance form, and improves the accuracy and robustness of posture recognition in more general human posture detection by using state-of-the-art techniques in 3D model reconstruction and human posture detection. Application opportunities have arisen.
Their research has advanced computer vision and 3D modeling techniques, impacting many fields such as healthcare, sports analysis, and animation. It will also enrich your knowledge about and the ability to preserve the rich cultural heritage of Bharatanatyam. We hope that all parties involved in this project will benefit from the research and that researchers in this field will be able to obtain near-perfect performance indicators. The evaluation shows how well the augmentation, preprocessing, and skeletonization work. Subsequent work will focus on validation and optimization to improve the pipeline's robustness and speed.
