AI Paper Atlas

Research map / 3D vision / 3D generation

Learning 3D Object Spatial Relationships from Pre-trained 2D Diffusion Models

Sangwon Baik, Hyeonwoo Kim, Hanbyul Joo

ICCV 2025 · 3D vision › 3D generation · approach: Score distillation for text-to-3D · 2 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 segmentation to PCA-only registration (ablation)OOR data generation · ParaHome OOR distributionsFrechet distance (lower better): 1.87 -> 1.43Improved
add PCA to segmentation-only registration (ablation)OOR data generation · ParaHome OOR distributionsFrechet distance (lower better): 1.50 -> 1.43Improved
use OOR diffusion instead of SceneTellerOOR diffusion · Pairwise OOR generationVLM Score: 64.67 -> 75.67Improved
use OOR diffusion instead of SceneTellerOOR diffusion · Pairwise OOR generationCLIP Score: 29.06 -> 29.11No clear effect
use OOR diffusion instead of SceneMotifCoderOOR diffusion · Pairwise OOR generationUser Study (%): 22.21 -> 54.02Improved
use OOR diffusion instead of GraphDreamerOOR diffusion · Multi-object OOR (original rendering)VLM Score: 2.50 -> 97.50Improved
use OOR diffusion instead of GraphDreamerOOR diffusion · Multi-object OOR (original rendering)User Study (%): 11.88 -> 88.12Improved
use OOR diffusion instead of GraphDreamerOOR diffusion · Multi-object OOR (Canny ControlNet)VQA Score: 0.44 -> 0.71Improved
use OOR diffusion instead of GraphDreamerOOR diffusion · Multi-object OOR (Depth ControlNet)CLIP Score: 28.51 -> 28.47No clear effect
use OOR diffusion instead of GraphDreamerOOR diffusion · Multi-object OOR (Depth ControlNet)VQA Score: 0.62 -> 0.79Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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