Research map / 3D vision / Face and shape
GARDEN: Gravity-Aligned Reconstruction of Disentangled ENvironments from RGB images
Jiahao Sun, Dingkun Wei, Zehong Shen, Hongyu Zhou, Yujun Shen, Liang Li
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 Gravity-View alignment (ablation) | GARDEN (point cloud) · NVS ablation Hypersim | RMSE: 0.2961 -> 0.2751 | Improved |
| add Gravity-View alignment (ablation) | GARDEN (point cloud, with simulation) · NVS ablation Hypersim | RMSE: 0.3002 -> 0.2751 | Improved |
| add Gravity-View alignment (ablation) | GARDEN (3DGS) · NVS ablation Hypersim | RMSE: 0.3171 -> 0.2840 | Improved |
| add Gravity-View alignment (ablation) | GARDEN (point cloud) · NVS ablation Hypersim | Completeness: 0.0743 -> 0.0387 | Improved |
| add physically grounded integration beyond GV plus FoundationPose (ablation) | GARDEN (point cloud) · NVS ablation Hypersim | RMSE: 0.2753 -> 0.2751 | Improved |
| factorize scene (objects plus background) instead of DA3 output (ablation) | GARDEN (point cloud) · NVS ablation Hypersim | Completeness: 0.0516 -> 0.0387 | Improved |
| factorize scene instead of DA3 output (ablation) | GARDEN (point cloud) · NVS ablation Hypersim | Accuracy: 0.0082 -> 0.0658 | Hurt |
| factorize scene instead of DA3 output (ablation) | GARDEN (3DGS) · NVS ablation Hypersim | RMSE: 0.3068 -> 0.2840 | Improved |
| replace CAD retrieval with gravity-aligned factorization (post-sim) | GARDEN vs LiteReality · Object-centric material recovery | RMSE: 0.2163 -> 0.1880 | Improved |
| replace CAD retrieval with factorization (post-sim) | GARDEN vs LiteReality · Object-centric material recovery | SSIM: 0.4353 -> 0.4240 | No clear effect |
| replace CAD retrieval with factorization (post-sim) | GARDEN vs LiteReality · Holistic scene | LPIPS: 0.6522 -> 0.4081 | Improved |
| evaluate static rather than post-simulation | GARDEN (point cloud) · Holistic scene | RMSE: 0.1616 -> 0.1511 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LiteReality — LiteReality: Graphic-Ready 3D Scene Reconstruction from RGB-D Scans (NeurIPS 2025) · better on 5 of 7 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
- PRISM: Programming Interactive Scenes from Monocular Images for Embodied Simulation (NeurIPS 2026 · similarity 0.86)
- Simulation-Ready Compositional 3D Scene Reconstruction from a Single Image (NeurIPS 2026 · similarity 0.85)
- DecoupledGaussian: Object-Scene Decoupling for Physics-Based Interaction (CVPR 2025 · similarity 0.85)
- HoGS: Unified Near and Far Object Reconstruction via Homogeneous Gaussian Splatting (CVPR 2025 · similarity 0.84)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.84)
- Horizon-GS: Unified 3D Gaussian Splatting for Large-Scale Aerial-to-Ground Scenes (CVPR 2025 · similarity 0.84)
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.
