AI Paper Atlas

Research map / Point clouds / Shape and surface completion

Ray-Path-Aware Virtual Point Removal on 2D Layer-Wise Nearest Point Map

DAEHYEON JEON, Kyungdon Joo, Jae-Young Sim

ECCV 2026 · Point clouds › Shape and surface completion · approach: Point cloud upsampling and normals · 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 Layer-Wise Nearest Point Map instead of naive interval-based projection (ablation)Ray-Path-Aware framework · UNIST LS3DPCF1: 0.2043 -> 0.7895Improved
use Ray-Path-Aware Attention instead of ViT (ablation)virtual point removal network · UNIST LS3DPCF1: 0.6533 -> 0.7895Improved
use Ray-Path-Aware Attention instead of UFormer (ablation)virtual point removal network · UNIST LS3DPCF1: 0.6290 -> 0.7895Improved
use Ray-Path-Aware Attention instead of H-win (ablation)virtual point removal network · UNIST LS3DPCprecision: 0.5331 -> 0.8194Improved
use Ray-Path-Aware Attention instead of H-win (ablation)virtual point removal network · UNIST LS3DPCrecall: 0.8654 -> 0.7618Hurt
use Ray-Path-Aware Attention instead of H-win+V-win (ablation)virtual point removal network · UNIST LS3DPCF1: 0.6794 -> 0.7895Improved
use 3D PCA normals instead of 2D range-map gradient normals (ablation)Ray-Path-Aware Attention · UNIST LS3DPCF1: 0.7705 -> 0.7895Improved
add uniform noise [0,3] to normals (ablation)Ray-Path-Aware Attention · UNIST LS3DPCF1: 0.7895 -> 0.7799Hurt
add uniform noise [0,9] to normals (ablation)Ray-Path-Aware Attention · UNIST LS3DPCF1: 0.7895 -> 0.7516Hurt
add 20% outlier normals (ablation)Ray-Path-Aware Attention · UNIST LS3DPCF1: 0.7895 -> 0.7306Hurt
fine-tune on 3DRN real data after synthetic pre-trainingProposed · UNIST LS3DPCF1: 0.7541 -> 0.7895Improved
fine-tune on UNIST LS3DPC after synthetic pre-trainingProposed · 3DRNF1: 0.7874 -> 0.8062Improved

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