Research map / 3D vision / 3D generation
Learning to Exploit Stability for 3D Scene Parsing
Yilun Du, Zhijian Liu, Hector Basevi, Ales Leonardis, Bill Freeman, Josh Tenenbaum, Jiajun Wu
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 physics stability supervision (ablation) | Factored3D (GT boxes) · SUNCG 0.1% labels | Avg IoU: 0.149 -> 0.164 | Improved |
| add physics stability supervision (ablation) | Factored3D (GT boxes) · SUNCG 10% labels | stability score SPhys: 0.31 -> 0.54 | Improved |
| add finetuning on unlabeled data with physics after physics supervision (ablation) | Factored3D + P · SUNCG 10% labels | Avg IoU: 0.261 -> 0.266 | No clear effect |
| add physics supervision when training on full SUNCG (ablation) | Prim R-CNN · SUNCG full data | mAP0.5: 0.402 -> 0.393 (mAP0.5), 0.672 -> 0.680 (mAP0.3) | Mixed |
| add physics supervision when training on full SUNCG (ablation) | Prim R-CNN · SUNCG full data | stability score SPhys: 0.42 -> 0.54 | Improved |
| add physics supervision when training on full SUNCG (ablation) | layout prediction · SUNCG full data | wall offset MSE: 0.299 -> 0.297 | Improved |
| add physics supervision to Factored3D on SceneNet RGB-D (ablation) | Factored3D · SceneNet RGB-D 100% | Avg IoU: 0.0276 -> 0.0304 | Improved |
| replace Factored3D with Edge Boxes by end-to-end Prim R-CNN | primitive prediction · SUNCG full data | mAP0.5: 0.046 -> 0.402 | Improved |
| finetune on unlabeled SUN RGB-D with physics supervision | Factored3D (GT boxes) · SUN RGB-D | 0.3 mAP: 0.016 -> 0.056 | Improved |
| finetune on unlabeled SUN RGB-D with physics supervision | Prim R-CNN · SUN RGB-D | 0.3 mAP: 0.024 -> 0.056 | Improved |
| add physics supervision to Factored3D with Prim R-CNN setting at 10% labels (ablation) | Factored3D · SUNCG 10% labels (Table 3) | mAP0.5: 0.303 -> 0.300 | No clear effect |
Closest papers by meaning
- Naturally Supervised 3D Visual Grounding with Language-Regularized Concept Learners (CVPR 2024 · similarity 0.86)
- SceneScaffold: Active Scene-State Construction for Unified 3D Scene Understanding (NeurIPS 2026 · similarity 0.86)
- Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding (NeurIPS 2024 · similarity 0.86)
- MLLMs Need 3D-Aware Representation Supervision for Scene Understanding (NeurIPS 2025 · similarity 0.86)
- SceneFun3D: Fine-Grained Functionality and Affordance Understanding in 3D Scenes (CVPR 2024 · similarity 0.86)
- Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers (NeurIPS 2024 · similarity 0.86)
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.
