AI Paper Atlas

Research map / 3D vision / 3D generation

SuperDec: 3D Scene Decomposition with Superquadrics Primitives

Elisabetta Fedele, Boyang Sun, Leonidas Guibas, Marc Pollefeys, Francis Engelmann

ICCV 2025 · 3D vision › 3D generation · approach: Learning 3D reconstruction from images

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
use SuperDec instead of SQ baselinesuperquadric decomposition · ShapeNet in-categoryL2 Chamfer: 0.279 -> 0.047Improved
use SuperDec instead of EMS optimizationsuperquadric decomposition · ShapeNet in-categoryL2 Chamfer: 1.345 -> 0.047Improved
use SuperDec instead of CSA cuboidsdecomposition · ShapeNet out-of-categoryL2 Chamfer: 0.427 -> 0.061Improved
use SuperDec instead of CSAdecomposition · ScanNet++ objectsL2 Chamfer: 0.41 -> 0.11Improved
use SuperDec instead of EMSdecomposition · ScanNet++ objectsnumber of primitives: 4.25 -> 5.18Hurt
use SuperDec instead of CSAdecomposition · Replica objectsL2 Chamfer: 0.70 -> 0.19Improved
use SuperDec instead of cuboids for path planningrepresentation · 15 ScanNet++ scenessuccess rate (%): 61.23 -> 91.71Improved
use SuperDec instead of dense point cloud for path planningrepresentation · 15 ScanNet++ scenesmemory (MB): 19.286 -> 0.042Improved
use SuperDec instead of occupancy grid for path planningrepresentation · 15 ScanNet++ scenessuccess rate (%): 100.00 -> 91.71Hurt
add existence head alpha at training and inference (ablation)SuperDec · ShapeNet in-categoryL2 Chamfer: 0.049 -> 0.047Mixed
use predicted alpha at inference (ablation)SuperDec · ShapeNet in-categoryL2 Chamfer: 0.046 -> 0.047No clear effect
use farthest point sampling instead of random (ablation)SuperDec · ScanNet++ objectsL2 Chamfer: 0.13 -> 0.11Improved

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