Research map / 3D vision / Human pose estimation
Learning 3D Part Assembly from a Single Image
Yichen Li, Kaichun Mo, Lin Shao, Minhyuk Sung, Leonidas Guibas
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 part-instance image segmentation module (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.363 -> 0.454 | Improved |
| add part-instance image segmentation module (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | assembly Chamfer distance: 0.084 -> 0.067 | Improved |
| add part-instance image segmentation module (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (invisible): 0.164 -> 0.270 | Improved |
| add L2 rotation loss (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.426 -> 0.454 | Improved |
| add both graph convolution stages (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.403 -> 0.454 | Improved |
| add second-stage graph convolution (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.434 -> 0.454 | Improved |
| add second-stage graph convolution (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (invisible): 0.239 -> 0.270 | Improved |
| add image feature to pose prediction (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.403 -> 0.454 | Improved |
| add global feature (ablation) | single-image 3D part assembly pipeline · PartNet Chair (Level-3) | part accuracy (total): 0.418 -> 0.454 | Improved |
| use proposed method instead of B-Global baseline (adds segmentation-based 2D-3D grounding and graph conv) | B-Global · PartNet Chair (Level-3) | part accuracy: 0.170 -> 0.454 | Improved |
| use proposed method instead of strongest baseline B-GRU | B-GRU · PartNet Chair (Level-3) | part accuracy: 0.310 -> 0.454 | Improved |
| use proposed method instead of B-GRU | B-GRU · PartNet Chair (Level-3) | assembly Chamfer distance: 0.107 -> 0.067 | Improved |
Closest papers by meaning
- Imagine: Image-Guided 3D Part Assembly with Structure Knowledge Graph (AAAI 2025 · similarity 0.91)
- Neural Assembler: Learning to Generate Fine-Grained Robotic Assembly Instructions from Multi-View Images (AAAI 2025 · similarity 0.91)
- Category-Level Multi-Part Multi-Joint 3D Shape Assembly (CVPR 2024 · similarity 0.89)
- Two by Two: Learning Multi-Task Pairwise Objects Assembly for Generalizable Robot Manipulation (CVPR 2025 · similarity 0.89)
- Generative 3D Part Assembly via Part-Whole-Hierarchy Message Passing (CVPR 2024 · similarity 0.89)
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.88)
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.
