AI Paper Atlas

Research map / 3D vision / Depth and stereo

Motion Basis Learning for Unsupervised Deep Homography Estimation With Subspace Projection

Nianjin Ye, Chuan Wang, Haoqiang Fan, Shuaicheng Liu

ICCV 2021 · 3D vision › Depth and stereo · approach: Dense feature matching for correspondence · 85 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 4-offset regression with homography flow basis weights (ablation)CA-Unsupervised with mask · CA-Unsupervised homography test (4.2k) ablation (avg)point matching error (lower is better): 0.88 -> 0.83Improved
add feature identity loss (FIL) (ablation)mask + basis network · CA-Unsupervised homography test (4.2k) ablation (avg)point matching error (lower is better): 0.83 -> 0.77Improved
replace mask module with LRR blocks (ablation)homography flow basis network · CA-Unsupervised homography test (4.2k) ablation (avg)point matching error (lower is better): 0.83 -> 0.60Improved
add feature identity loss (FIL) (ablation)LRR + basis network · CA-Unsupervised homography test (4.2k) ablation (avg)point matching error (lower is better): 0.60 -> 0.50Improved
add feature extractor f() with 3 conv layers instead of photometric loss only (no FIL) (ablation)homography estimator · CA-Unsupervised homography test, feature extractor depthpoint matching error (lower is better): 0.76 -> 0.60Improved
increase f() from 2 to 3 layers without FIL (ablation)homography estimator · CA-Unsupervised homography test, feature extractor depthpoint matching error (lower is better): 0.61 -> 0.60No clear effect
add FIL to 3-layer f() training (ablation)homography estimator · CA-Unsupervised homography test, feature extractor depthpoint matching error (lower is better): 0.60 -> 0.50Improved
use full BasesHomo instead of strongest feature-based pipeline (SuperPoint + SuperGlue + MAGSAC)BasesHomo · CA-Unsupervised homography test (4.2k) (avg over 5 scene types)point matching error: 0.63 -> 0.50Improved

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