VR-Based System for Pyrotechnics Skill Transfer and CNN-Based Determination of Gunpowder Mixing Ratios to Achieve Target Firework Colors

Despite being an important part of traditional culture, pyrotechnics as a profession is on the decline in Japan. The purpose of this study was to try to arrest this trend by enabling effective skill transfer from pyrotechnicians to beginners and enthusiasts using the latest information technology. To this end, we created a virtual reality (VR)-based system where users can experience the fireworks manufacturing process in a virtual space, and learn the skills in a hands-on fashion using a VR glove and goggles. We also implemented a deep learning-based regression model to determine mixing ratios of gunpowder to achieve target firework colors. Two methods were proposed to create training and testing datasets—one yielded 90% accuracy for the regression model, while the other yielded more than 60%. The proposed framework is expected to make pyrotechnical skills and experimentation with new fireworks designs more accessible to non-experts and promote the popularity of the profession among the youth.