A YOLO-GAN Integrated Framework for Real-Time Opera Stage Performance Assistance and Motion Capture Feedback

Abstract

The system offers technical support for stage performance innovation and digital preservation of traditional dance, enabling real-time interaction between performers and stage effects. Experiments validate its utility in enhancing performance immersion and heritage documentation. The proposed system combines remote sensing images with high spatial and temporal resolution and a convolutional neural network (CNN) for feature extraction and change detection of urban boundaries. An improved YOLO model is adopted to enhance the accuracy and real-time performance of detecting buildings and urban areas. Real-time motion capture technology (200 Hz sampling) tracks performer movements, providing 3D pose data for stage effect synchronization. This enhances detection of dynamic actions, with 98.7% joint precision in skeletal tracking. The improved YOLO model achieves 52.6% recognition accuracy in dynamic stage environments, with a 73.4% enhancement in detection speed compared to baseline models. In complex backgrounds, accuracy reaches 88.9%, while simpler scenarios yield 45.2%. This study presents a real-time stage performance assistance system integrating YOLO, motion capture, and GAN technologies. The system leverages YOLO for high-speed performer detection (82.5% accuracy in dynamic scenes) and motion capture for 3D pose tracking (98.7% joint precision). A Pix2Pix GAN generates adaptive stage backgrounds (realism score 4.3/5), enabling interactive feedback for lighting and sound effects. Experiments show 12 ms response latency and 85.6% system stability in live performances, demonstrating its utility for digital protection of dance intangible cultural heritage.