A team of researchers at Anna University in Chennai has accurately identified and categorized 108 foundational Bharatanatyam dance poses using advanced computational techniques.

J. Jayanthi and P. U. Maheswari wanted to capture the essence of the Indian classical dance form through its rich history and complex movements, as described in the ancient Sanskrit text Natyashastra, a treatise on the performing arts. This, the researchers say, will be crucial for preserving and teaching the dance form.

The team captured all the 108 dance poses depicted in stone statues at the tenth century Chidambaram Nataraja temple near Chennai. They also collected publicly available images of various Bharatanatyam poses and processed them through skeletonization – reducing the poses to their essential structures – and data augmentation. They examined the images using a convolutional neural network model to classify the dance poses.

The researchers used the MediaPipe library for body key point detection and the deep learning architecture Inception-ResNet-v2 for pose classification. This approach improved the accuracy of pose identification and helped them digitally preserve the dance form. They reconstructed 3D models of the dance poses to provide deeper understanding of their anatomical intricacies and movement patterns.

The model can be used to classify several dance poses and forms with high accuracy. It also opens avenues for virtual reality experiments, robotics, and human-computer interaction, the researchers say.