AI Paper Atlas

Research map / Detection and segmentation / CNN features

Pixel-wise Planarity for High-Precision Monocular Plane Segmentation

Ahmetcan Yavuz, Alpay Ozkan, Rémi Pautrat, Shaohui Liu, Marc Pollefeys

ECCV 2026 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 0 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 relative depth cue to normal cue in region growing (ablation)region growing · ScanNet++precision @0.1cm: 43.0 -> 45.5Improved
add predicted planarity map as third gating cue (ablation)region growing · ScanNet++precision @0.1cm: 45.5 -> 62.1Improved
add predicted planarity gating (recall cost) (ablation)region growing · ScanNet++recall @0.5cm: 64.2 -> 60.1Hurt
use depth cue alone instead of normals (ablation)region growing · ScanNet++precision @0.1cm: 0.6 vs 43.0Hurt
source normals and depth from different pretrained models instead of MoGeV2 (ablation)pipeline · ScanNet++precision @0.1cm: 62.1 -> 27.3Hurt
replace MoGeV2 normals and depth with Metric3D (ablation)pipeline · ScanNet++precision @0.1cm: 62.1 -> 21.0Hurt
use Focal loss instead of BCE for planarity head (ablation)planarity head · ScanNet++precision @0.1cm: 62.1 -> 57.9Mixed
use Dice loss instead of BCE for planarity head (ablation)planarity head · ScanNet++precision @0.1cm: 62.1 -> 62.3No clear effect
use shallow conv probe instead of standalone planarity head (ablation)planarity head · ScanNet++precision @0.1cm: 62.1 -> 69.0Mixed
retrain ZeroPlane on our GT with same backboneZeroplane · ScanNet++precision @0.1cm: 16.6 vs 62.1Hurt
tighten region-growing thresholds (default to strict) (ablation)region growing · ScanNet++precision @0.1cm / recall: 68.0 -> 81.9; recall 42.3 -> 32.3Mixed
use planarity-guided method instead of Zeroplane on out-of-domain outdoor dataOurs · VKITTI2 (trained indoors)precision @2cm: 22.5 -> 62.1Improved

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