Planar Instance 3D Reconstruction Leveraging Planar Structural Priors

ICRA2026 Anonymous Authors

PIPS reconstructs planar instance meshes, achieving alignment between geometry and semantics through planar structural priors.

Abstract

Teaser
Planar structures, ubiquitous in man-made indoor environments, enable compact and accurate scene abstraction for various downstream tasks. Recent methods distill planar features into learning-based MVS geometries to obtain coherent 3D plane estimation from multi-view inputs. However, the lack of explicit planar instance definitions hinders semantic–geometry alignment, leading to distorted geometry and mismatched semantics. To address this, we propose PIPS, a planar instance 3D reconstruction method that leverages planar structural priors for both structure-guided planar segmentation (SGPS module) and multi-view planar integration (MVPI module). The planar instance point clouds are regularized by planar distances and then converted into complete planar meshes via a planar-aware ball-pivoting strategy. Extensive experiments on hundreds of indoor scenes demonstrate the superior performance of our method, which is less dependent on annotations and requires no feature optimization. The effectiveness of each component is further verified through comprehensive ablation studies.

ScanNetv2

Reference

Choose a result from

Ours

Reference

Choose a result from

Ours

ScanNet++

Reference

Choose a result from

Ours

Reference

Choose a result from

Ours