Pith. sign in

REVIEW 3 cited by

Compact 3D Gaussian Representation for Radiance Field

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2311.13681 v2 pith:5YCEJBTB submitted 2023-11-22 cs.CV cs.GR

Compact 3D Gaussian Representation for Radiance Field

classification cs.CV cs.GR
keywords representationgaussianrenderinghighnumberperformanceachievingattributes
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
abstract

Neural Radiance Fields (NeRFs) have demonstrated remarkable potential in capturing complex 3D scenes with high fidelity. However, one persistent challenge that hinders the widespread adoption of NeRFs is the computational bottleneck due to the volumetric rendering. On the other hand, 3D Gaussian splatting (3DGS) has recently emerged as an alternative representation that leverages a 3D Gaussisan-based representation and adopts the rasterization pipeline to render the images rather than volumetric rendering, achieving very fast rendering speed and promising image quality. However, a significant drawback arises as 3DGS entails a substantial number of 3D Gaussians to maintain the high fidelity of the rendered images, which requires a large amount of memory and storage. To address this critical issue, we place a specific emphasis on two key objectives: reducing the number of Gaussian points without sacrificing performance and compressing the Gaussian attributes, such as view-dependent color and covariance. To this end, we propose a learnable mask strategy that significantly reduces the number of Gaussians while preserving high performance. In addition, we propose a compact but effective representation of view-dependent color by employing a grid-based neural field rather than relying on spherical harmonics. Finally, we learn codebooks to compactly represent the geometric attributes of Gaussian by vector quantization. With model compression techniques such as quantization and entropy coding, we consistently show over 25$\times$ reduced storage and enhanced rendering speed, while maintaining the quality of the scene representation, compared to 3DGS. Our work provides a comprehensive framework for 3D scene representation, achieving high performance, fast training, compactness, and real-time rendering. Our project page is available at https://maincold2.github.io/c3dgs/.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. GETA-3DGS: Automatic Joint Structured Pruning and Quantization for 3D Gaussian Splatting

    cs.LG 2026-05 unverdicted novelty 7.0

    GETA-3DGS is the first automatic joint structured pruning and quantization framework for 3D Gaussian Splatting, achieving roughly 5x storage reduction on standard datasets without per-scene thresholds.

  2. 3D Gaussian Splatting for Scientific Particle Data Compression and Rendering

    cs.GR 2026-07 conditional novelty 6.0

    ParticleGS uses 3D Gaussian splats to mimic ParaView renderings of 281M-particle data, reaching 30 dB PSNR at 65x compression and rendering at 662 FPS.

  3. Compact 3D Gaussian Splatting For Dense Visual SLAM

    cs.CV 2024-03 unverdicted novelty 6.0

    A compact 3D Gaussian Splatting SLAM system reduces Gaussian count and parameter size via masking and a geometry codebook while preserving SOTA reconstruction quality and pose accuracy.