NVGS: Neural Visibility for Occlusion Culling in 3D Gaussian Splatting

February 20, 2026

Reading time: 14 minute

...

📝 Original Info

Title: NVGS: Neural Visibility for Occlusion Culling in 3D Gaussian Splatting
ArXiv ID: 2511.19202
Date: 2023-10-15
Authors: Yuanming Hu, Zexiang Xu, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavalli, Yao Zhou, Kalyan Sunkavalli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zexiang Xu, Kalyan Sunkavlli, Yao Zhou, Kalyan Sunkavlli, Zex산…

📝 Abstract

https://brent-zoomers.github.io/nvgs/ Figure 1. Shows a scene composed of multiple, separately trained 3DGS assets. Top left shows the gsplat implementation of 3DGS[28], Top right V3DG [27] and Bottom Ours. Our approach uses significantly less VRAM, consistently achieves higher image quality, and increases FPS. We achieve this by combining a novel instanced rasterizer and our neural visibility MLP, which enables occlusion culling.

💡 Deep Analysis

📄 Full Content

3D Gaussian Splatting [5] (3DGS) has proven itself to be a valuable tool for 3D reconstruction in recent years, due to its ability to represent scenes with great accuracy while also facilitating both fast training and rendering times. As 3DGS transitions from research to actual applications, the focus on efficiently rendering scenes, which are increasing in both scale and complexity, becomes more important. Previous methods [6,9,19,21,27] have followed this line of thinking and introduced the traditional Level of Detail (LoD) approach to the context of 3DGS. This enables larger scenes to be trained and rendered by utilizing hierarchical techniques to select the currently relevant parts of a scene. Other widely used approaches in graphics, such as occlusion culling, however, are not a natural fit due to the semi-transparent nature of 3D Gaussians. In graphics, these techniques typically rely on the use of opaque triangles, for which determining visibility is straightforward.

Our results show that standard volume rendering implicitly encodes a soft form of occlusion: once transmittance saturates, subsequent splats can be discarded without any significant loss in color. This allows us to determine Gaus-Figure 2. Left shows the donut asset from the front. Right we show the donut as if rendered from the front from a different viewpoint. We then show the Gaussians that our approach culls in red, along with what we would optimally render.

sians that are not visible to the current viewpoint, which, similar to backface culling for meshes, are located mainly on the backside of the object, as shown in Fig. 2. Another nice property of this formulation is that it scales with distance (Fig. 3). As more primitives are mapped to the same set of pixels, more Gaussians can be discarded, even those lying on the front of the object. Capturing this information, however, requires a large amount of information to be stored. In this work, we propose an efficient pipeline that captures and utilizes visibility information during rendering to discard Gaussians, thereby avoiding the need for expensive preprocessing in the rasterization pipeline. We also propose a novel instanced rasterizer specifically tailored for composed scenes consisting of multiple 3DGS assets. Gaussians are only instantiated after passing through multiple culling phases, which drastically impacts both VRAM usage and rendering speed. Baking the visibility data into a lightweight MLP that is fully integrated into this rasterizer results in a highly optimized pipeline for composed scenes. Similar to their complementary use in graphics, our occlusion culling approach shows complementary properties with LoD, opening up avenues for further optimization of large-scale scenes. In summary, our contributions are:

• We propose a novel approach for occlusion culling in 3DGS rendering that is based on compressing the viewdependent visibility functions of all Gaussians into a small MLP. • We provide an efficient, virtualized, and instanced rendering pipeline for 3DGS capable of processing scenes over 100 million Gaussians at real-time frame rates. • We demonstrate that neural network queries can be integrated into the renderer to minimize excessive VRAM usage and global memory traffic, thereby improving performance.

In combination, these contributions outperform the current state of the art for composed scenes in terms of VRAM usage, image quality, and frame rate for short, medium, and, in some cases, long distances.

3DGS [5] has gained immense popularity within the community ever since its introduction. Despite its success, 3DGS still had some limitations, some of which have been addressed in follow-up work, such as its memory footprint and popping artifacts. To reduce the memory footprint, several approaches have been proposed that compress the final trained scene [2,13,15], utilizing codebooks or exploiting the redundant nature within the Gaussian parameters. Other works, such as StopThePop [18], have focused on solving the popping artefacts resulting from global sorting. The densification step used in 3DGS to promote exploration also inspired several follow-up works [7,20]. These works propose alternative strategies for splitting, cloning, or simply moving Gaussians, enabling higher-quality scene reconstructions. In our work, we focus on a different issue related to compression: redundancy at render time. By baking the visibility of Gaussians into a lightweight MLP, we can efficiently avoid expensive preprocessing operations on Gaussians that will not be used during rendering.

Within 3DGS, a large degree of redundancy exists, as evidenced by the extensive work on pruning and compression.

One common trend among these methods is their focus on redundancy at a global level, meaning that Gaussians that do not contribute to the scene will be pruned. Zhang et al. [30] learns this masking by using the Gimbal-Sigmoid method to achieve an approximate binary mask, which can

📄 Read Full PDF on ArXiv