Research map / 3D vision / 3D generation
InSpace: Structure-Aware 3D Indoor Scene Generation from a Single 360° Image
Gwanhyeong Koo, Hyunsu Kim, Youngji Kim, Taejae Lee, Siwoo Lim, Sunjae Yoon, Suyong Yeon, Chang D. Yoo
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add view-selective cross-attention to the coarse-structure model (ablation) | InSpace Stage 2 · ERP-FRONT test (Stage 2) | Voxel IoU / CD: IoU 44.36 -> 57.54; CD 15.92 -> 7.68 | Improved |
| add view-selective cross-attention (ablation) | InSpace Stage 2 · ERP-FRONT test (Stage 2) | F1: 55.61 -> 69.07 | Improved |
| add VS attention for robustness to camera distance and room size (ablation) | InSpace Stage 2 · ERP-FRONT test (varying camera position, room area) | Voxel IoU, CD | Improved |
| add Layout-Guided Structure Inversion at t0=0.5 (vs pure noise start) (ablation) | InSpace Stage 2 · ERP-FRONT test (Stage 2) | Voxel IoU: 57.54 -> 58.48 | Improved |
| add Layout-Guided Structure Inversion at t0=0.7 (default) (ablation) | InSpace Stage 2 · ERP-FRONT test (Stage 2) | Voxel IoU: 57.54 -> 57.09 | No clear effect |
| add Layout-Guided Structure Inversion at t0=0.7 (ablation) | InSpace Stage 3 · ERP-FRONT test (Stage 3 scene) | Chamfer Distance x10^3: 1.52 -> 0.63 | Improved |
| add Layout-Guided Structure Inversion at t0=0.7 (ablation) | InSpace Stage 3 · ERP-FRONT test (Stage 3 asset) | F1@.01: 53.42 -> 56.71 | Improved |
| raise inversion t0 from 0.3 to 0.7 (ablation) | InSpace · ERP-FRONT test (Stage 2 vs Stage 3) | IoU (Stage 2) vs mesh metrics (Stage 3): Stage 2 IoU 58.06 -> 57.09; Stage 3 scene CD 1.48 -> 0.63 | Mixed |
| add view-selective and asset-selective attention with global scene self-attention (full InSpace) vs perspective-image single-image methods | InSpace · 20 ERP-FRONT test scenes, asset level | CD, F-score | Improved |
Closest papers by meaning
- SceneCraft: Layout-Guided 3D Scene Generation (NeurIPS 2024 · similarity 0.89)
- OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization (NeurIPS 2025 · similarity 0.88)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.87)
- S-INF: Towards Realistic Indoor Scene Synthesis via Scene Implicit Neural Field (AAAI 2025 · similarity 0.87)
- DeBaRA: Denoising-Based 3D Room Arrangement Generation (NeurIPS 2024 · similarity 0.87)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.87)
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.
