AI Paper Atlas

Research map / 3D vision / Gaussian splatting

3D Gaussian Splatting at Arbitrary Resolutions with Compact Proxy Anchors

Mingyun Jeong, Seongro Yoon, Francois Bremond, Donghyeon Cho

CVPR 2026 · 3D vision › Gaussian splatting · approach: Compact and efficient Gaussian splatting

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add FiLM resolution-embedding instead of none (ablation)Scaffold-GS-based method · Mip-NeRF360 (average PSNR)PSNR: 27.199 -> 28.183Improved
use FiLM modulation instead of simple additive resolution-embedding (ablation)Scaffold-GS-based method · Mip-NeRF360 (average PSNR)PSNR: 28.036 -> 28.183Improved
add additive resolution-embedding vs none (ablation)Scaffold-GS-based method · Tanks&Temples (average SSIM)SSIM: 0.856 -> 0.866Improved
use FiLM instead of additive embedding (ablation)Scaffold-GS-based method · Tanks&Temples (average SSIM)SSIM: 0.866 -> 0.870Improved
apply pixel coverage gate only at test time (post-hoc) on Scaffold-GS (ablation)Scaffold-GS · DeepBlending (average PSNR)PSNR: 29.465 -> 28.147Hurt
train the pixel coverage gate jointly instead of post-hoc (ablation)Scaffold-GS-based method · DeepBlending (average PSNR)PSNR: 28.147 -> 29.905Improved
add pixel coverage gate (jointly trained) (ablation)Scaffold-GS-based method · Mip-NeRF360 (bonsai, stump)Gaussian size distribution / blurImproved
add resolution-adaptive generation and proxy anchors to Scaffold-GSScaffold-GS · Mip-NeRF360 (10% resolution)PSNR: 24.681 -> 29.742Improved
add resolution-adaptive generation and proxy anchors to Scaffold-GSScaffold-GS · Mip-NeRF360 (100% resolution)PSNR: 26.498 -> 26.710Improved
store proxy anchors with residual anchor predictor instead of all anchorsScaffold-GS · Mip-NeRF360storage (MEM): 78.9MB -> 41.6MBImproved
proxy anchor storage compared with Context-GS and TC-GS compressionOurs vs anchor compression methods · Mip-NeRF360storage (MEM): 8.2MB (TC-GS) vs 41.6MB (Ours)Hurt
use proxy anchors with fewer feature-retaining anchorsScaffold-GS · Tanks&Templesnumber of anchors: 208.0K -> 67.2KImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy