SOTAVerified

Surface Normals Estimation

Surface normal estimation deals with the task of predicting the surface orientation of the objects present inside a scene. Refer to Designing Deep Networks for Surface Normal Estimation (Wang et al.) to get a good overview of several design choices that led to the development of a CNN-based surface normal estimator.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–25 of 39 papers

TitleStatusHype
Learning from the Giants: A Practical Approach to Underwater Depth and Surface Normals Estimation—0
Fine-Tuning Image-Conditional Diffusion Models is Easier than You ThinkCode4
PolyMaX: General Dense Prediction with Mask TransformerCode0
Stanford-ORB: A Real-World 3D Object Inverse Rendering BenchmarkCode1
Large-scale Monocular Depth Estimation in the Wild—0
MSECNet: Accurate and Robust Normal Estimation for 3D Point Clouds by Multi-Scale Edge ConditioningCode1
MIMIC: Masked Image Modeling with Image CorrespondencesCode1
Neural-PBIR Reconstruction of Shape, Material, and Illumination—0
iDisc: Internal Discretization for Monocular Depth EstimationCode3
NeFII: Inverse Rendering for Reflectance Decomposition with Near-Field Indirect IlluminationCode1
CORE: Co-planarity Regularized Monocular Geometry Estimation with Weak Supervision—0
NeAF: Learning Neural Angle Fields for Point Normal EstimationCode1
HSurf-Net: Normal Estimation for 3D Point Clouds by Learning Hyper SurfacesCode1
GraphFit: Learning Multi-scale Graph-Convolutional Representation for Point Cloud Normal EstimationCode1
Shape, Light, and Material Decomposition from Images using Monte Carlo Rendering and DenoisingCode2
GRIT: General Robust Image Task BenchmarkCode1
InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene UnderstandingCode2
Extracting Triangular 3D Models, Materials, and Lighting From ImagesCode2
Estimating and Exploiting the Aleatoric Uncertainty in Surface Normal EstimationCode1
AdaFit: Rethinking Learning-based Normal Estimation on Point CloudsCode1
NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown IlluminationCode1
PhySG: Inverse Rendering with Spherical Gaussians for Physics-based Material Editing and Relighting—0
NeRD: Neural Reflectance Decomposition from Image CollectionsCode1
How Well Do Self-Supervised Models Transfer?Code1
On Deep Learning Techniques to Boost Monocular Depth Estimation for Autonomous Navigation—0
Show:102550
← PrevPage 1 of 2Next →
All datasetsPCPNetStanford-ORBNYU-Depth V2ScanNetV2IBims-1NYU-Depth V2 Surface NormalsPASCAL ContextTaskonomy

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1Nesti-NetRMSE 12.41—Unverified
2Iter-NetRMSE 11.84—Unverified
3DeepFitRMSE 11.8—Unverified
4AdaFitRMSE 10.76—Unverified
5GraphFitRMSE 10.26—Unverified
6NeAFRMSE 10.22—Unverified
7HsurfRMSE 10.11—Unverified
8MSECNetRMSE 9.76—Unverified
#ModelMetricClaimedVerifiedStatus
1NeRFactorCosine Distance0.29—Unverified
2NeRDCosine Distance0.28—Unverified
3PhySGCosine Distance0.17—Unverified
4NVDiffRecCosine Distance0.06—Unverified
5Neural-PBIRCosine Distance0.06—Unverified
6InvRenderCosine Distance0.06—Unverified
7NVDiffRecMCCosine Distance0.04—Unverified
#ModelMetricClaimedVerifiedStatus
1Metric3Dv2(L, FT)% < 11.2568.8—Unverified
2PolyMaX(ConvNeXt-L)% < 11.2565.66—Unverified
3iDisc% < 11.2563.8—Unverified
4Bae et al.% < 11.2562.2—Unverified
5Marigold + E2E FT(zero-shot)% < 11.2561.4—Unverified
6Floors are Flat% < 11.2559.5—Unverified
#ModelMetricClaimedVerifiedStatus
1Metric3Dv2 (g2, In-domain)% < 11.2577.8—Unverified
2Bae et al.% < 11.2571.1—Unverified
3Floors are Flat% < 11.2550.9—Unverified
#ModelMetricClaimedVerifiedStatus
1Marigold + E2E FT(zero-shot)% < 11.2569.9—Unverified
2Metric3Dv2(g2, ZS)% < 11.2569.7—Unverified
#ModelMetricClaimedVerifiedStatus
1DSNRMSE12.2—Unverified
#ModelMetricClaimedVerifiedStatus
1InvPTMean Angle Error14.15—Unverified
#ModelMetricClaimedVerifiedStatus
1X-TC (Cross-Task Consistency)L1 error4.8—Unverified