GATED TRANSFORMER WITH GRAPH NEURAL NETWORKS AND TEMPORAL CONVOLUTIONAL NETWORKS FOR DANCE CHOREOGRAPHY

Abstract

Dance choreography is an expressive and dynamic art form that demands intricate coordination between body movements and music. Although traditional approaches to dance choreography generation, such as those based on Bi-LSTM, tend to encounter difficulties in capturing nuanced details and temporal dependencies. In this study, we introduce an innovative approach that employs the power of Gated Transformer, Graph Neural Networks (GNN), and Temporal Convolutional Networks (TCN) to transform dance choreography generation. Our proposed model not only captures the intricacy of dance movements but also guarantees accurate synchronization with music, presenting new possibilities for creative choreography. With the use of our approach, choreographers will be able to produce more complex and creative dance routines that are in sync with the music and are capable of capturing the nuances of the art form. By combining the power of Gated Transformer, Graph Neural Networks (GNN), and Temporal Convolutional Networks (TCN), our approach promises to revolutionize the way dance choreography is generated.