AI Paper Atlas

Research map / Point clouds / Shape and surface completion

NoiseSDF2NoiseSDF: Learning Clean Neural Fields from Noisy Supervision

Tengkai Wang, Weihao Li, Ruikai Cui, Shi Qiu, Nick Barnes

ICML 2026 · Point clouds › Shape and surface completion · approach: Distance-function surface reconstruction

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
train with noisy SDF targets (NoiseSDF2NoiseSDF) on top of 3DS2V3DS2V · ShapeNet (sigma=0.01)IoU: 0.884 -> 0.929Improved
train with noisy SDF targets on top of 3DS2V at higher noise3DS2V · ShapeNet (sigma=0.02)Normal Consistency: 0.733 -> 0.868Improved
train with noisy SDF targets under Laplace noise3DS2V · ShapeNet Chair Laplace (sigma=0.01)IoU: 0.909 -> 0.908No clear effect
train with noisy SDF targets under Gaussian noise with mean 0.023DS2V · ShapeNet Chair Gaussian (mu=0.02)F-Score: 0.549 -> 0.524Hurt
use noisy targets in place of clean SDF targets (ablation)NoiseSDF2NoiseSDF · ShapeNet Chair (3DS2V network)IoU: 0.939 (clean) vs 0.927 (noisy)No clear effect
use 3DILG instead of 3DS2V to generate noisy targets (ablation)NoiseSDF2NoiseSDF · ShapeNet Chair (3DS2V network)IoU: 0.927 -> 0.903Hurt
add cross-attention block to fine-tuned FC layer (ablation)denoising network · ShapeNet ChairIoU: 0.884 -> 0.922Improved
fine-tune the whole decoder instead of FC+CA (ablation)denoising network · ShapeNet ChairIoU: 0.922 -> 0.927Improved
randomly initialize the whole network instead of 3DS2V init (ablation)denoising network · ShapeNet ChairIoU: 0.927 -> 0.929No clear effect
average three noisy targets (1-to-3) instead of one (ablation)NoiseSDF2NoiseSDF · ShapeNet ChairIoU: 0.927 -> 0.923No clear effect
add test-time optimization with 200 noisy samplesNoiseSDF2NoiseSDF · Famous medium noiseChamfer: 0.0160 -> 0.0108Improved
use end-to-end noisy-SDF training instead of denoise-then-reconstructNoiseSDF2NoiseSDF · ShapeNet ChairIoU: 0.882 -> 0.927Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 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