Research map / 3D vision / Face and shape
Manual-PA: Learning 3D Part Assembly from Instruction Diagrams
Jiahao Zhang, Anoop Cherian, Cristian Rodriguez, Weijian Deng, Stephen Gould
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 assembly order as soft guidance (vs w/o Order) (ablation) | Manual-PA · PartNet Chair | PA: 79.07 -> 89.06 | Improved |
| add part assembly order as soft guidance (vs w/o Order) (ablation) | Manual-PA · PartNet Table | PA: 74.03 -> 87.41 | Improved |
| add part assembly order as soft guidance (vs w/o Order) (ablation) | Manual-PA · PartNet Storage | PA: 48.54 -> 56.51 | Improved |
| add sequential order information to part features (PE on parts) (ablation) | Manual-PA · PartNet Chair, GT order | PA: 72.22 -> 79.31 | Improved |
| add step-diagram positional encoding with correct correspondence to part PE (ablation) | Manual-PA · PartNet Chair, GT order | PA: 79.31 -> 95.38 | Improved |
| use PE instead of OneHot encoding for step and part (cross-attention) (ablation) | Manual-PA · PartNet Chair, GT order | PA: 81.22 -> 95.38 | Improved |
| use PE instead of OneHot for part only (self-attention) (ablation) | Manual-PA · PartNet Chair, GT order | PA: 79.31 vs 79.07 | No clear effect |
| add batch-level sampling in contrastive permutation learning (ablation) | Manual-PA · PartNet Chair | PA: 56.93 -> 89.06 | Improved |
| add batch-level sampling in contrastive permutation learning (ablation) | Manual-PA · PartNet Chair | KT: 0.410 -> 0.789 | Improved |
| add Sinkhorn normalization to similarity matrix (ablation) | Manual-PA · PartNet Chair | PA: 89.06 -> 88.25 | No clear effect |
| replace learned order with ground-truth order (ablation) | Manual-PA · PartNet Chair | PA: 89.06 -> 95.38 | Improved |
| use manual diagrams instead of single image (Image-PA) | Manual-PA · PartNet Chair | PA: 62.67 -> 89.06 | Improved |
Closest papers by meaning
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.90)
- Imagine: Image-Guided 3D Part Assembly with Structure Knowledge Graph (AAAI 2025 · similarity 0.87)
- Neural Assembler: Learning to Generate Fine-Grained Robotic Assembly Instructions from Multi-View Images (AAAI 2025 · similarity 0.87)
- Two by Two: Learning Multi-Task Pairwise Objects Assembly for Generalizable Robot Manipulation (CVPR 2025 · similarity 0.87)
- Category-Level Multi-Part Multi-Joint 3D Shape Assembly (CVPR 2024 · similarity 0.87)
- Generative 3D Part Assembly via Part-Whole-Hierarchy Message Passing (CVPR 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.
