Neural Radiance Fields (NeRF) for Dynamic Lighting Effects and Real-Time Rendering Optimization in Virtual Fashion Show Garment Design

Abstract

Due to the presence of many physical fashion stores and virtual fashion exhibitions on the internet, virtual clothing research is becoming increasingly common. Rendering fabrics with photorealistic quality under dynamic lighting and its practical application remain the main obstacles that must be overcome. Based on the new technology of neural radiance fields, a method for real-time digital rendering of very delicate clothing is proposed. By encoding the characteristics of clothing materials through a special network structure, and then using dynamic lighting techniques to achieve precise rendering of appearance and fabric light response. A more optimized real-time rendering pipeline, using GPU acceleration and adaptive sampling techniques, ensures smooth operation of clothing in terms of image quality. Based on various environmental lighting and dynamic changes in similar clothing, as well as a complete dataset containing real and hypothetical clothing items. Compared to traditional physics-based rendering methods and previous neural rendering methods, it improves image clarity, structural consistency, and perceptual effects. Based on this paper, the framework can support interactive e-commerce, digital fashion design, and virtual fitting programs. In the future, various complex textiles and lighting variations will be added. Provide a reliable and scalable system design to achieve the visualization of high-quality digital garment images.