AI Paper Atlas

Research map / 3D vision / Depth and stereo

An Analysis of SVD for Deep Rotation Estimation

Jake Levinson, Carlos Esteves, Kefan Chen, Noah Snavely, Angjoo Kanazawa, Afshin Rostamizadeh, Ameesh Makadia

NeurIPS 2020 · 3D vision › Depth and stereo · approach: Camera model and calibration for SfM · 41 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
replace 6D Gram-Schmidt projection with SVD orthogonalization (SVD-Train)rotation regression network · Point cloud alignment (supervised)mean geodesic error (deg): 2.24 -> 1.63Improved
replace 6D Gram-Schmidt projection with SVD orthogonalizationrotation regression network · Self-supervised point cloud alignmentmean geodesic error (deg): 2.39 -> 1.58Improved
replace 6D with SVD-Trainrotation regression network · ModelNet10 sofa posemean geodesic error (deg): 20.25 -> 18.01Improved
replace 6D with SVD-Trainrotation regression network · Human pose inverse kinematicsmean joint error (cm): 1.70 -> 1.61Improved
apply SVD projection during training (SVD-Train) rather than only at inference (SVD-Inf) (ablation)SVD orthogonalization · Point cloud alignment (supervised)mean geodesic error (deg): 2.64 -> 1.63Improved
apply SVD projection during training rather than only at inference (ablation)SVD orthogonalization · ModelNet10 chair posemean geodesic error (deg): 21.38 -> 21.25No clear effect
apply SVD projection during training rather than only at inference (ablation)SVD orthogonalization · Pascal3D+ mean over 12 categoriesaccuracy@10deg (%): 39.9 -> 35.1Hurt
replace SVD-Train with S2-Reg on Pascal3D+ (reverse comparison)SVD-Train · Pascal3D+ mean over 12 categoriesmedian error (deg): 11.3 vs 14.6Hurt
use SVD-Train instead of Euler anglesKITTI depth/ego-motion model · KITTI depthSq Rel: 3.163 -> 2.517Improved
use SVD-Train instead of axis-angleKITTI depth model · KITTI depthAbs Rel: 0.208 -> 0.209Mixed
train with geodesic loss instead of squared Frobenius loss (ablation)SVD-Train · Point cloud alignmentmean geodesic error (deg): 1.63 -> 2.05No clear effect
add learning rate decay (ablation)point cloud alignment training · Point cloud alignmentmean geodesic error (deg)No clear effect

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