Research map / 3D vision / 3D generation
Structured3D: A Large Photo-realistic Dataset for Structured 3D Modeling
Jia Zheng, Junfei Zhang, Jing Li, Rui Tang, Shenghua Gao, Zihan Zhou
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 synthetic data with real data (s + r) instead of real only | HorizonNet · PanoContext (layout estimation) | 3D IoU (%): 83.42 -> 84.45 | Improved |
| pre-train on synthetic then fine-tune (s -> r) instead of real only | HorizonNet · 2D-3D-S (layout estimation) | 3D IoU (%): 84.33 -> 86.01 | Improved |
| pre-train on synthetic then fine-tune (s -> r) instead of real only | LayoutNet · PanoContext (layout estimation) | 3D IoU (%): 84.15 -> 84.77 | Improved |
| pre-train on synthetic then fine-tune (s -> r) instead of real only | LayoutNet · 2D-3D-S (layout estimation) | 3D IoU (%): 83.39 -> 84.04 | Improved |
| add synthetic data with real data via BGC (s + r) instead of real only | LayoutNet · PanoContext (layout estimation) | 3D IoU (%): 84.15 -> 84.96 | Improved |
| add synthetic data with real data via BGC (s + r) instead of real only | HorizonNet · 2D-3D-S (layout estimation) | 3D IoU (%): 84.33 -> 84.36 | No clear effect |
| increase synthetic pre-training images from 1k to 10k (ablation) | HorizonNet · PanoContext (layout estimation) | 3D IoU (%): 83.77 -> 85.27 | Improved |
| increase synthetic pre-training images from 1k to 10k (ablation) | LayoutNet · PanoContext (layout estimation) | 3D IoU (%): 83.81 -> 84.77 | Improved |
| increase synthetic pre-training images from 1k to 5k (ablation) | LayoutNet · 2D-3D-S (layout estimation) | 3D IoU (%): 83.57 -> 84.55 | Improved |
| increase synthetic pre-training images from 5k to 10k (ablation) | HorizonNet · 2D-3D-S (layout estimation) | 3D IoU (%): 86.35 -> 86.01 | Hurt |
| add discriminator to align layout outputs between synthetic and real domains | LayoutNet · 2D-3D-S (layout estimation) | 3D IoU (%): 57.18 -> 70.08 | Improved |
| add discriminator to align layout outputs between synthetic and real domains | LayoutNet · PanoContext (layout estimation) | 3D IoU (%): 75.64 -> 76.91 | Improved |
Closest papers by meaning
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.90)
- 3D-FRONT: 3D Furnished Rooms With layOuts and semaNTics (ICCV 2021 · similarity 0.90)
- Photo3D: Advancing Photorealistic 3D Generation through Structure-Aligned Detail Enhancement (CVPR 2026 · similarity 0.90)
- Estimating Generic 3D Room Structures from 2D Annotations (NeurIPS 2023 · similarity 0.89)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.88)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (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.
