Digital Intelligence Technology-Driven Transformation and Innovation Research on Choreography Design

Abstract With the development of the times and the progress of science and technology, stage design has also gained more development, and at the same time, it also promotes the stage design more and more tends to be intelligent and creative, intelligent accompanied by creativity almost at the same time existence and development, the two are complementary. This paper applies Stable Diffusion technology to generate the tone and style of choreography design, and adjusts the parameters of the generated image through LoRA model to realize the style fine-tuning. Attention mechanism is introduced on the basis of SD model to enhance the representation of object embedding vectors. Use the scene layout network based on region convolutional network to improve the accuracy and stability of target detection. Introducing the three concepts of Query, Key and Value, a CLIP text encoder is designed so that the sublayer network generates clearer images. The stage performers are detected and recognized by YOLO algorithm, and the Nearest Neighbor Rule Classification and Matching algorithm is applied to calculate the distance between each sample, extract the human body key point features, and realize the stage light tracking. The physiological characteristic signals of the audience were collected by using eye-tracking instrument, and the spatial elements of the choreography that were paid more attention by the subjects were the spatial interface elements and the set elements, which were more eye-catching and attractive than the other four types of spatial elements, and the total duration of the gaze was 104.624 s and 63.552 s. The correct rate of the direction of the rotational tracking of the luminaires controlled by the KNN and RANSAC algorithms used in this paper reaches 97%, and the rotation degree error is between [-18.72%,22.3%]. It meets the stage lighting tracking performance requirements.