Bharatanatyam dancing stances are classified via machine learning. Using state-of-the-art computer techniques, a group of researchers from Anna University in Chennai have accurately recognized and categorized 108 basic Bharatanatyam dance positions. AI is also capable of modeling and conserving other conventional performing arts.
This research delves into the specialized area of human action detection, with a focus on identifying Indian classical dance positions, particularly those used in Bharatanatyam. According to the Natyashastra, a “Karana” in dance is a synchronized and melodic movement of the body, hands, and feet.
Karana incorporates sthaana (body postures), chaari (leg movements), and nritta hasta (hand movements). The intricate stone carvings that adorn the Nataraj temples in Chidambaram depict the 108 karanas that the Natyashastra specifies, notwithstanding their vast number. These carvings also show how Lord Shiva is related to these movements. Because there are so many distinct hand and body postures, mudras (hand gestures), facial expressions, and head motions in Bharatanatyam, it is challenging to automate pose detection.
To make this difficult task easier, automation and image processing techniques are used. The proposed method consists of four steps: feature extraction from images, deep learning network-based convolution neural network model (InceptionResNetV2) classification of dance postures, mesh creation from point clouds for 3D model visualization, and skeletonization and data augmentation techniques for image acquisition and preprocessing.
Identification is made easier by utilizing state-of-the-art technologies like body key point recognition via the MediaPipe library and deep learning networks. A critical step called data augmentation expands small datasets to boost the accuracy of the model. Analysis and interpretation were made simpler by the convolution neural network model’s effective recognition of intricate dancing moves. This innovative approach creates a standard for increasing effectiveness and accessibility for scholars and practitioners of Indian classical dance, while also making Bharatanatyam posture recognition easier.
Human posture detection has proven to be extremely challenging in computer vision because of its numerous and diverse applications in daily life. Because posture identification may have an impact on a person’s health, it is crucial in the context of Indian classical dance, particularly Bharatanatyam.
InceptionResNetV2, a novel convolutional neural network model based on a deep learning network, is presented by the authors of this research. This model accurately classifies 108 dance positions and builds upon key features found in MediaPipe. After a thorough analysis of the published literature on the subject, their strategy was developed.
Their method relies on separately removing spatial and depth components from the images, then combining the two sets of data to identify different positions. This unique approach helps their design distinguish different positions more successfully, as initially suggested in their technique and subsequently verified by comparisons and analysis of the study’s results.
Furthermore, their suggested design can accommodate different placements because of their feature extraction approach. Future research endeavors will mostly focus on optimizing performance via hyperparameter adjustments.
Lastly, the ongoing efforts to identify Indian classical dance postures, particularly in Bharatanatyam, have benefited immensely from their work. By utilizing state-of-the-art techniques for 3D model reconstruction and human pose identification, their research has improved the accuracy and robustness of posture recognition in this intricate dance style and opened up possibilities for more widespread applications in human position detection.
Their research improved 3D modeling and computer vision techniques, which have applications in animation, sports analysis, healthcare, and other domains. It also improves our understanding of and capacity to protect Bharatanatyam’s rich cultural heritage. Everyone involved in this study will benefit from their work, which they believe will point researchers in this direction toward nearly perfect performance measures. The evaluation demonstrates the effectiveness of augmentation, preprocessing, and skeletonization; the next study focuses on optimization and validation for increased pipeline robustness and performance.
