AI Paper Atlas

Research map / 3D vision / 3D generation

Diffusion-based dual-view reflection removal

Tianyi Xu, Zifeng Wang, Boyang Lyu, Shuchen Weng, Boxin Shi

ECCV 2026 · 3D vision › 3D generation · approach: Diffusion-based novel view synthesis · 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 camera pose encoding (ablation)dual-view diffusion · DualRef-VPSNR: 23.53 -> 24.17Improved
add camera pose encoding (ablation)dual-view diffusion · DualRef-RPSNR: 27.47 -> 28.19Improved
use two distinct views instead of one view (ablation)dual-view diffusion w/o pose · DualRef-VPSNR: 21.59 -> 23.53Improved
use two distinct views instead of identical inputs (ablation)dual-view diffusion w/o pose · DualRef-RPSNR: 26.12 -> 27.47Improved
use frame-aware spatial attention instead of RAFT flow warping (ablation)SD2.1 dual-view · DualRef-VPSNR: 21.60 -> 24.17Improved
use frame-aware spatial attention instead of Cat3D-style joint attention (ablation)dual-view diffusion · DualRef-VPSNR: 22.09 -> 24.17Improved
use frame-aware spatial attention instead of 3D ResNet (SIR-style) (ablation)dual-view diffusion · DualRef-VPSNR: 20.87 -> 24.17Improved
use channel-wise fusion with attention instead of width concatenation (ablation)dual-view diffusion · DualRef-RPSNR: 25.83 -> 28.19Improved
use frame-aware spatial attention instead of Cat3D-style joint attention (ablation)dual-view diffusion · DualRef-RLPIPS: 0.1244 -> 0.0983Improved
fine-tune DSIT on DualRef training dataDSIT · DualRef-VPSNR: 19.54 -> 20.55Improved
fine-tune ZF24 on DualRef training dataZF24 · DualRef-VPSNR: 19.80 -> 20.17Improved
increase number of input views beyond two (ablation)dual-view diffusion · DualRef-VqualityNo clear effect

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