AI Paper Atlas

Research map / 3D vision / Gaussian splatting

Efficient Scene Modeling via Structure-Aware and Region-Prioritized 3D Gaussians

Guangchi Fang, Bing Wang

TPAMI 2025 · 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 blur split to 3DGS (ablation)3DGS · Mip-NeRF360 (ablation)LPIPS: 0.216 -> 0.195Improved
add depth reinitialization after blur split (ablation)3DGS + Blur Split · Mip-NeRF360 (ablation)LPIPS: 0.195 -> 0.175Improved
add blur split to 3DGS (ablation)3DGS · Mip-NeRF360 (ablation)PSNR: 27.47 -> 27.47No clear effect
reinitialize with blending depth instead of midpoint depth (ablation)depth reinitialization · Mip-NeRF360 (ablation)PSNR: 27.54 -> 17.67Hurt
use Gaussian center instead of midpoint depth (ablation)depth reinitialization · Mip-NeRF360 (ablation)PSNR: 27.54 -> 27.57No clear effect
add aggressive model growth to Mini-Splatting* (ablation)Mini-Splatting* · Mip-NeRF360 (ablation)training time: 7min59s -> 4min4sImproved
add occluded Gaussian culling after aggressive growth (ablation)MSv2 · Mip-NeRF360 (ablation)training time: 4min4s -> 3min34sImproved
add aggressive model growth to Mini-Splatting* (ablation)Mini-Splatting* · Mip-NeRF360 (ablation)PSNR: 27.23 -> 27.43Improved
add occluded Gaussian culling (ablation)MSv2 · Mip-NeRF360 (ablation)PSNR: 27.43 -> 27.33Hurt
use 3K-iteration shortened densification instead of full aggressive growth (ablation)MSv2 · Mip-NeRF360 (ablation)PSNR: 27.33 -> 26.49Hurt
use lower gradient threshold instead of full aggressive growth (ablation)MSv2 · Mip-NeRF360 (ablation)PSNR: 27.33 -> 26.58Hurt
use only Agg-Depth instead of full aggressive growth (ablation)MSv2 · Mip-NeRF360 (ablation)PSNR: 27.33 -> 27.12Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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