Splat the Net: Radiance Fields with Splattable Neural Primitives

2025-10-09

AI Summary

To address the trade-off between expressiveness and efficiency in neural rendering versus primitive-based rendering, this paper introduces Splatting-Ready Neural Primitives (SRNPs): shallow neural networks parameterizing bounded density fields, enabling the first analytical, view-consistent splatting kernel computation without ray marching. Our method tightly integrates the representational power of neural radiance fields with the rasterization efficiency of point-based splatting, introducing learnable, large-scale, adaptive neural primitives that drastically reduce both primitive count and model parameters—without requiring complex control mechanisms. Experiments on novel-view synthesis demonstrate that SRNPs achieve rendering quality and frame rates comparable to 3D Gaussian Splatting, while using only 1/10 the number of primitives and 1/6 the total parameters. The approach thus delivers high fidelity, real-time performance, and exceptional model compactness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Neuro-Symbolic Learning

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Security and Privacy: Privacy-enhancing technologies

Abstract

Radiance fields have emerged as a predominant representation for modeling 3D scene appearance. Neural formulations such as Neural Radiance Fields provide high expressivity but require costly ray marching for rendering, whereas primitive-based methods such as 3D Gaussian Splatting offer real-time efficiency through splatting, yet at the expense of representational power. Inspired by advances in both these directions, we introduce splattable neural primitives, a new volumetric representation that reconciles the expressivity of neural models with the efficiency of primitive-based splatting. Each primitive encodes a bounded neural density field parameterized by a shallow neural network. Our formulation admits an exact analytical solution for line integrals, enabling efficient computation of perspectively accurate splatting kernels. As a result, our representation supports integration along view rays without the need for costly ray marching. The primitives flexibly adapt to scene geometry and, being larger than prior analytic primitives, reduce the number required per scene. On novel-view synthesis benchmarks, our approach matches the quality and speed of 3D Gaussian Splatting while using $10 imes$ fewer primitives and $6 imes$ fewer parameters. These advantages arise directly from the representation itself, without reliance on complex control or adaptation frameworks. The project page is https://vcai.mpi-inf.mpg.de/projects/SplatNet/.

Problem

• Reconciles neural radiance field expressivity with splatting efficiency
• Enables analytical line integrals without costly ray marching
• Reduces primitive count and parameters while maintaining quality

Innovation

• Splattable neural primitives combine neural expressivity with splatting efficiency
• Each primitive encodes neural density field with shallow network
• Analytical line integrals enable efficient splatting without ray marching