On the Impacts of Spherical Harmonics and Gaussian Counts in WebGL-Based 3-D Gaussian Splatting

3D Gaussian Splatting (3DGS) is an emerging technique for training and representing photorealistic real-world 3D scenes. Due to its high rendering speed (e.g., over 100 frames-per-second on a CUDA-enabled GPU), it has been viewed as a game-changing representation for immersive media. One popular application of 3DGS is streaming and rendering 3D Gaussians directly in modern browsers via WebGL. The large number of Gaussians in a trained 3DGS scene and their attributes, particularly the spherical harmonics for view-dependent color, makes optimizing WebGL rendering efficiency a significant challenge. This paper describes a WebGL-based web viewer that supports rendering 3DGS scenes with different spherical harmonics (SH) levels. We systematically evaluate the impact of SH levels and Gaussian count on rendering performance, in terms of visual quality and frame rates. Our framework provides researchers with a platform to explore optimization techniques for efficient web-based 3DGS rendering.