AI Paper Atlas

Research map / 3D vision / Gaussian splatting

SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization

Jiun Lee, Jaekwang Kim, Sangmin Lee

ECCV 2026 · 3D vision › Gaussian splatting · approach: Compact and efficient Gaussian splatting · 0 citations (OpenAlex)

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 sub-pixel Gaussian reparameterizer (ablation)MVSplat · RealEstate10K K=4PSNR: 19.46 -> 25.15Improved
Add feature aggregation (deformable attention) on top of SPGR (ablation)SubSplat · RealEstate10K K=16PSNR: 19.49 -> 22.65Improved
Add feature aggregation on top of SPGR (ablation)SubSplat · RealEstate10K K=4PSNR: 25.15 -> 25.52Improved
Use three deformable-attention stages instead of one (ablation)SubSplat · RealEstate10KPSNR: 23.94 -> 25.52Improved
Use SPGR instead of uniform increase of Gaussian count (ablation)MVSplat · RealEstate10K K=4PSNR: 20.13 -> 25.52Improved
Increase depth residual scale beta_z from 0.25 to 1.00 (ablation)SubSplat · RealEstate10KPSNR: 25.52 -> 24.14Hurt
Use bounded color gain [0.75,1.25] vs none (ablation)SubSplat · RealEstate10KPSNR: 24.89 -> 25.52Improved
Use 3D sub-pixel densification instead of bilinear post upsamplingMVSplat · RealEstate10K 512PSNR: 24.01 -> 25.52Improved
Use 256 input with SPGR instead of 512 input baselinesSubSplat · RealEstate10K 512 outputlatency: 131 ms -> 42 ms vs MVSplat; 1.57 GB vs 4.27 GBImproved
Use SubSplat instead of PixelSplat at 1024x1024PixelSplat · RealEstate10K 1024PSNR: 21.44 -> 22.65Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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