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 Added Most Hyped Most Active Needs Verification Most Verified

Showing 21–30 of 39 papers

Title	Date	Tasks	Status	Hype
NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination	Jun 3, 2021	Depth PredictionImage Relighting	CodeCode Available	1
Stanford-ORB: A Real-World 3D Object Inverse Rendering Benchmark	Oct 24, 2023	Depth PredictionImage Relighting	CodeCode Available	1
AdaFit: Rethinking Learning-based Normal Estimation on Point Clouds	Aug 12, 2021	Surface Normals Estimation	CodeCode Available	1
SharinGAN: Combining Synthetic and Real Data for Unsupervised Geometry Estimation	Jun 7, 2020	Depth EstimationMonocular Depth Estimation	CodeCode Available	1
Large-scale Monocular Depth Estimation in the Wild	Sep 18, 2023	Depth EstimationDepth Prediction	—Unverified	0
CORE: Co-planarity Regularized Monocular Geometry Estimation with Weak Supervision	Jan 1, 2023	3D geometryDepth Estimation	—Unverified	0
Learning from the Giants: A Practical Approach to Underwater Depth and Surface Normals Estimation	Oct 2, 2024	3D ReconstructionAutonomous Navigation	—Unverified	0
Neural-PBIR Reconstruction of Shape, Material, and Illumination	Apr 26, 2023	Depth PredictionImage Relighting	—Unverified	0
On Deep Learning Techniques to Boost Monocular Depth Estimation for Autonomous Navigation	Oct 13, 2020	Autonomous NavigationAutonomous Vehicles	—Unverified	0
PhySG: Inverse Rendering with Spherical Gaussians for Physics-based Material Editing and Relighting	Apr 1, 2021	Depth PredictionImage Relighting	—Unverified	0

Show:10 25 50

← PrevPage 3 of 4Next →

All datasets PCPNet Stanford-ORB NYU-Depth V2 ScanNetV2 IBims-1 NYU-Depth V2 Surface Normals PASCAL Context Taskonomy

Benchmark Results

#	Model	Metric	Claimed	Verified	Status
1	Nesti-Net	RMSE	12.41	—	Unverified
2	Iter-Net	RMSE	11.84	—	Unverified
3	DeepFit	RMSE	11.8	—	Unverified
4	AdaFit	RMSE	10.76	—	Unverified
5	GraphFit	RMSE	10.26	—	Unverified
6	NeAF	RMSE	10.22	—	Unverified
7	Hsurf	RMSE	10.11	—	Unverified
8	MSECNet	RMSE	9.76	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	NeRFactor	Cosine Distance	0.29	—	Unverified
2	NeRD	Cosine Distance	0.28	—	Unverified
3	PhySG	Cosine Distance	0.17	—	Unverified
4	InvRender	Cosine Distance	0.06	—	Unverified
5	NVDiffRec	Cosine Distance	0.06	—	Unverified
6	Neural-PBIR	Cosine Distance	0.06	—	Unverified
7	NVDiffRecMC	Cosine Distance	0.04	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	Metric3Dv2(L, FT)	% < 11.25	68.8	—	Unverified
2	PolyMaX(ConvNeXt-L)	% < 11.25	65.66	—	Unverified
3	iDisc	% < 11.25	63.8	—	Unverified
4	Bae et al.	% < 11.25	62.2	—	Unverified
5	Marigold + E2E FT(zero-shot)	% < 11.25	61.4	—	Unverified
6	Floors are Flat	% < 11.25	59.5	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	Metric3Dv2 (g2, In-domain)	% < 11.25	77.8	—	Unverified
2	Bae et al.	% < 11.25	71.1	—	Unverified
3	Floors are Flat	% < 11.25	50.9	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	Marigold + E2E FT(zero-shot)	% < 11.25	69.9	—	Unverified
2	Metric3Dv2(g2, ZS)	% < 11.25	69.7	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	DSN	RMSE	12.2	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	InvPT	Mean Angle Error	14.15	—	Unverified

#	Model	Metric	Claimed	Verified	Status
1	X-TC (Cross-Task Consistency)	L1 error	4.8	—	Unverified