AI Paper Atlas

Research map / 3D vision / Face and shape

NeuS2: Fast Learning of Neural Implicit Surfaces for Multi-view Reconstruction

Yiming Wang, Qin Han, Marc Habermann, Kostas Daniilidis, Christian Theobalt, Lingjie Liu

ICCV 2023 · 3D vision › Face and shape · approach: Neural implicit surface reconstruction · 333 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 progressive coarse-to-fine hash encoding training (ablation)NeuS2 · DTUChamfer distance: 0.75 -> 0.70Improved
add Eikonal loss (ablation)NeuS2 · DTUChamfer distance: 1.48 -> 0.70Improved
add global transformation prediction and progressive training (ablation)NeuS2 dynamic · Dynamic ablationLPIPS: 0.0595 -> 0.0474Improved
add global transformation prediction on top of progressive training (ablation)NeuS2 dynamic · Dynamic ablationPSNR: 27.72 -> 28.18Improved
add progressive training only (remove GTP) (ablation)NeuS2 dynamic · Dynamic ablationPSNR: 27.78 -> 27.72No clear effect
replace finite-difference second-order derivatives with analytic ReLU-MLP formulation (ablation)NeuS2 · PyTorch comparisonPSNR: 33.4 -> 35.5Improved
use tailored second-order derivative formula instead of PyTorch autograd (ablation)NeuS2 · PyTorch comparisonruntime: 10 min -> 7 minImproved
use hash-encoded SDF with Eikonal loss instead of NeuS MLPNeuS · DTUruntime: 8 h -> 5 minImproved
use NeuS2 instead of Instant-NGP for geometry (surface constraints)Instant-NGP · DTUChamfer distance: 1.84 -> 0.70Improved
use NeuS2 instead of Instant-NGP for view synthesisInstant-NGP · DTUPSNR: 28.86 -> 28.82No clear effect
use NeuS2 incremental training instead of D-NeRF on real dynamic scenesD-NeRF · Real dynamic scenes D1PSNR: 21.48 -> 26.41Improved

Later papers that compared with this method

14 later accepted papers report NeuS2 as a baseline in their results tables; 11 of them report a better number on most of the results they share with it.

6 more not listed.

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