AI Paper Atlas

Research map / 3D vision / NeRF and radiance fields

OmniRen: Neural Rendering wih Heterogeneous Scene Primitives

Chong Zeng, Yue Dong, Pieter Peers, lvmin zhang, Maneesh Agrawala

ECCV 2026 · 3D vision › NeRF and radiance fields · approach: Inverse rendering with neural fields

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 attention sink to windowed attention (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 27.36 -> 28.25Improved
add sliding window attention to attention sink (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 26.92 -> 28.25Improved
add light source tokens to attention sink (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 27.40 -> 28.25Improved
add light source tokens to attention sink (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)LPIPS: 0.1081 -> 0.0997Mixed
add summarization tokens to attention sink (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 27.09 -> 28.25Improved
add both light and summary tokens to sink (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 26.13 -> 28.25Improved
use sparse attention instead of full attention (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 27.91 -> 28.25Improved
fewer summarization tokens (#seq/128 vs default) (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 28.25 -> 26.48Hurt
more summarization tokens (#seq/32 vs default) (ablation)RenderFormer-V2 · Synthetic random scenes, 16k primitives (stage 4)PSNR: 28.25 -> 27.78No clear effect
sparse attention vs RenderFormer brute-force attention at 83.2k trianglesRenderFormer-V2 · Large scenes (qualitative)image qualityImproved

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