AI Paper Atlas

Research map / 3D vision / 3D generation

ROOTS: Object-Centric Representation and Rendering of 3D Scenes

Chang Chen, Fei Deng, Sungjin Ahn

JMLR 2021 · 3D vision › 3D generation · approach: Learning 3D reconstruction from images · 27 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
use object-centric 3D models (ROOTS) instead of scene-level GQNGQN · 1-3 ShapesMSE: 40.62 -> 30.60Improved
use object-centric 3D models instead of scene-level GQNGQN · 3-5 ShapesMSE: 62.73 -> 54.45Improved
use object-centric 3D models instead of scene-level GQNGQN · Multi-Shepard-MetzlerMSE: 46.22 -> 42.22Improved
use object-centric 3D models instead of scene-level GQNGQN · ShapeNet arrangementMSE: 301.62 -> 212.77Improved
use ROOTS 3D segmentation instead of 2D IODINEIODINE · 1-3 ShapesARI: 0.8217 -> 0.9477Improved
use ROOTS instead of IODINE on Multi-Shepard-MetzlerIODINE · Multi-Shepard-MetzlerARI: Failed -> 0.9303Improved
use ROOTS instead of GQN when test scenes have fewer objects than trainingGQN · 1-3 Shapes (trained on 2-4 Shapes)MSE: 30.35 -> 24.27Improved
use ROOTS instead of GQN when test scenes have more objects than trainingGQN · 3-5 Shapes (trained on 2-4 Shapes)MSE: 86.68 -> 67.98Improved
use ROOTS representation instead of GQN for retrieve object taskGQN · Retrieve Object 3Daccuracy: 81.31% -> 90.38%Improved
use ROOTS representation instead of GQN for pair-wise relational reasoningGQN · Find Pairaccuracy: 12.48% -> 84.70%Improved
add more context observationsROOTS · Shapes counting and APobject counting accuracy and APImproved
add object-attention grouping before inferring object appearance in the encoder (ablation)ROOTS encoder · Shapestrainability: failed to decompose scenes -> trainableImproved

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