Development of Interactive Systems for Digital Art Exhibitions Integrating Neural Radiance Fields

Abstract

To foster a new paradigm for immersive digital art exhibitions that integrate virtual and real interactions, this paper designs and develops an interactive system integrating Neural Radiance Fields (NeRF), enabling realistic 3D visualization of textile designs and historical garments. As high-fidelity scene reconstruction and interactive rendering are increasingly important for electromagnetic information visualization, digital twins, and intelligent sensing applications, the proposed framework provides a practical reference for multimodal spatial representation and high-dimensional field reconstruction. The system effectively addresses the core challenges of dynamic representation of textile artworks and insufficient real-time interactivity through an interactive perception layer, a core processing layer, and a rendering and display layer. Users provide gesture, touch position, and pressure signals via a high-precision capacitive touchscreen to drive physically consistent interactions. The core innovation combines hash-encoded NeRF with a physically aware deformation network through an end-to-end differentiable joint optimization architecture, tightly coupling textile mechanical properties with high-dimensional visual representations to achieve consistency between geometric deformation and optical rendering. A hierarchical query mechanism and adaptive update strategy further enable real-time closed-loop interaction from user input to high-fidelity image generation. Experimental results demonstrate a rendering latency of 41.9 ms ± 9.8 ms with 23.9 ± 2.1 FPS under dynamic scaling, while achieving LPIPS of 0.31 ± 0.06 and SSIM of 0.75 ± 0.04 under high wind loads, providing both visual realism and interactive smoothness for immersive digital exhibition systems.