Dance Action Generation Model Based on Recurrent Neural Network

In recent years, modeling methods based on Deep Learning and Recurrent Neural Network (RNN) have achieved rapid development in various tasks in the field of computer vision, and video generation technology has also achieved remarkable results on this basis, with boosting effect. This paper analyzes the requirements and system design of the dance generation system based on cyclic neural network and clearly puts forward the specific functional requirements and nonfunctional requirements. Design, test each function of the system, and show the graphs generated by our application during training and testing, and visualize the specific value of the loss function during training. Combining the latest image generation technology and the most effective open source gesture detection project, a dance generation algorithm is completed, which can generate a video of the target person completing the specified dance action.