AI Paper Atlas

Research map / Point clouds / Shape and surface completion

Routing on Demand: DSNet for Efficient Progressive Point Cloud Denoising

Xiaoqian Cheng, Dong Xiao, Husen Li, Zheng Liu, Renjie Chen

CVPR 2026 · Point clouds › Shape and surface completion · approach: Point cloud upsampling and normals

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
replace fixed progressive path (IterativePFN) with dynamic per-patch routingIterativePFN · PU-Net 50K points, 2.5% noiseChamfer Distance (x10^4): 1.015 -> 0.762Improved
dynamic path selection versus static sequential path (4 stages) (ablation)DSNet-4 · PU-Net 50K points, 2.5% noiseChamfer Distance (x10^4): 1.017 -> 0.762Improved
dynamic path selection versus static sequential path (4 stages) (ablation)DSNet-4 · PU-Net 10K points, 2.5% noiseChamfer Distance (x10^4): 2.824 -> 2.604Improved
dynamic path selection versus static path at low noise (4 stages, 10K, 1%) (ablation)DSNet-4 · PU-Net 10K points, 1% noiseChamfer Distance (x10^4): 1.897 -> 1.829Improved
dynamic versus static path with 3 stages (ablation)DSNet-3 · PU-Net 10K points, 2% noise and 50K points, 2.5% noiseChamfer Distance (x10^4): 2.539 vs 2.521 (static) at 10K 2%; 0.845 vs 0.811 (static) at 50K 2.5%Mixed
increase number of denoising stages from 2 to 4 (ablation)DSNet · PU-Net 10K points, 2.5% noiseChamfer Distance (x10^4): 2.756 (L=2) -> 2.723 (L=3) -> 2.604 (L=4)Improved
increase number of denoising stages from 4 to 5 (ablation)DSNet · PU-Net 10K points, 2.5% noiseChamfer Distance (x10^4): 2.604 -> 2.635Hurt
DSNet on real Kinect scans versus best baselineDSNet · Kinect (real scans)Chamfer Distance (x10^4): 1.040 vs 0.974 (P2P-Bridge)No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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