Research map / 3D vision / Human pose estimation
SceneMaker: Open-set 3D Scene Generation with Decoupled De-occlusion and Pose Estimation Model
Yukai Shi, Weiyu Li, Zihao Wang, Hongyang Li, Xingyu Chen, Ping Tan, Lei Zhang
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 global self-attention (GSA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | CD-S: 0.0340 -> 0.0242 | Improved |
| add global self-attention (GSA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | F-Score-O: 0.6293 -> 0.8121 | Improved |
| add local self-attention (LSA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | CD-O: 0.0901 -> 0.0294 | Improved |
| add local self-attention (LSA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | IoU-B: 0.7733 -> 0.7555 | Mixed |
| add decoupled local cross-attention (LCA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | CD-O: 0.0429 -> 0.0294 | Improved |
| add decoupled local cross-attention (LCA) (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | IoU-B: 0.7882 -> 0.7555 | Mixed |
| give complete point clouds instead of single-image estimate (ablation) | SceneMaker pose model · Pose estimation ablation (GT meshes, 3D-Front, trained from scratch) | CD-S: 0.0242 -> 0.0064 | Improved |
| add 200K open-set scene dataset (ablation) | SceneMaker · Open-set test (1K scenes) | CD-S: 0.1538 -> 0.0285 | Improved |
| add 200K open-set scene dataset (ablation) | SceneMaker · 3D-Front indoor test (1K scenes, severe occlusion) | CD-S: 0.0381 -> 0.0470 | Hurt |
| fine-tune image editing model on 10K de-occlusion dataset (vs Flux Kontext) | De-occlusion model · De-occlusion validation (1K images, 500 classes) | PSNR: 13.91 -> 15.03 | Improved |
| decouple de-occlusion from 3D generation (vs Amodal3R) | SceneMaker · Occluded object generation (3D-Front renders) | Chamfer distance: 0.0443 -> 0.0409 | Improved |
| SceneMaker vs MIDI on MIDI test set | SceneMaker · MIDI test set (1K scenes) | CD-S: 0.080 -> 0.051 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- MIDI — MIDI: Multi-Instance Diffusion for Single Image to 3D Scene Generation (CVPR 2025) · better on 4 of 4 shared results
- Amodal3R — Amodal3R: Amodal 3D Reconstruction from Occluded 2D Images (ICCV 2025) · better on 3 of 3 shared results
- PartCrafter — PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025) · better on 2 of 2 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
- From Programs to Poses: Factored Real-World Scene Generation via Learned Program Libraries (NeurIPS 2025 · similarity 0.86)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.86)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.86)
- PartGen: Part-level 3D Generation and Reconstruction with Multi-view Diffusion Models (CVPR 2025 · similarity 0.85)
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.85)
- LEAP: Liberate Sparse-View 3D Modeling from Camera Poses (ICLR 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.
