Research map / Detection and segmentation / Open-set and few-shot detection
Physically-Based Rendering for Indoor Scene Understanding Using Convolutional Neural Networks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Pretrain with physically based indoor/outdoor-lit renders before NYUv2 finetuning | Normal estimation · NYUv2 | mean angular error (°): 23.38 -> 21.74 | Helped |
| Add indoor lighting to physically based rendering (ablation) | Normal estimation · NYUv2 | mean angular error (°): 49.33 -> 28.82 | Helped |
| Add synthetic-image selection (ablation) | Normal estimation with MLT-IL/OL pretraining · NYUv2 | mean angular error (°): 28.82 -> 27.90 | Helped |
| Use physically based indoor/outdoor-lit renders instead of OpenGL indoor-light renders (ablation) | Normal estimation without finetuning · NYUv2 | mean angular error (°): 33.06 -> 28.82 | Helped |
| Add real-scene context rather than training on isolated objects (ablation) | Normal estimation · NYUv2 | mean angular error (°): 48.78 -> 27.90 | Helped |
| Pretrain on physically based rendered RGB images | Semantic segmentation · NYUv2 | mean IoU: 31.7 -> 33.2 | Helped |
| Use physically based rendering instead of OpenGL rendering (ablation) | Semantic segmentation · NYUv2 | mean IoU: 32.8 -> 33.2 | Helped |
| Use color input instead of HHA depth input (ablation) | Semantic segmentation with synthetic pretraining · NYUv2 | mean IoU: 30.2 -> 33.2 | Helped |
| Pretrain on physically based renders before NYUv2 finetuning | Object boundary detection · NYUv2 | OSD: 0.713 -> 0.725 | Helped |
| Use physically based rendering instead of OpenGL indoor-light rendering (ablation) | Object boundary detection with NYUv2 finetuning · NYUv2 | OSD: 0.716 -> 0.725 | Helped |
| Finetune the synthetic-pretrained model on NYUv2 (ablation) | Object boundary detection with MLT-IL/OL pretraining · NYUv2 | OSD: 0.604 -> 0.725 | Helped |
Closest papers by meaning
- InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior (ICLR 2024 · similarity 0.86)
- DORSal: Diffusion for Object-centric Representations of Scenes $\textit{et al.}$ (ICLR 2024 · similarity 0.86)
- S-INF: Towards Realistic Indoor Scene Synthesis via Scene Implicit Neural Field (AAAI 2025 · similarity 0.86)
- SpatialLM: Training Large Language Models for Structured Indoor Modeling (NeurIPS 2025 · similarity 0.86)
- InternScenes: A Large-scale Simulatable Indoor Scene Dataset with Realistic Layouts (NeurIPS 2025 · similarity 0.86)
- Habitat Synthetic Scenes Dataset (HSSD-200): An Analysis of 3D Scene Scale and Realism Tradeoffs for ObjectGoal Navigation (CVPR 2024 · similarity 0.85)
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.