Research map / 3D vision / 3D generation
IKEA-Manual: Seeing Shape Assembly Step by Step
Ruocheng Wang, Yunzhi Zhang, Jiayuan Mao, Ran Zhang, Chin-Yi Cheng, Jiajun Wu
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 |
|---|---|---|---|
| Recursively group parts by geometric similarity rather than connect all parts in one step | Assembly plan generation · IKEA-Manual | Hard matching F1: 10.78 -> 16.30 | Improved |
| Recursively group parts by geometric similarity rather than connect all parts in one step | Assembly plan generation · IKEA-Manual | Simple matching precision: 100 -> 44.90 | Hurt |
| Use geometric similarity alone to build assembly-plan groups | GeoCluster · IKEA-Manual | Hard matching F1: 16.30 | Hurt |
| Fine-tune the pretrained segmentation model on IKEA-Manual examples | PointNet-conditioned U-Net · IKEA-Manual part-conditioned segmentation test | IoU: 25.31 | Mixed |
| Optimize pose from masks using differentiable rendering instead of learned pose prediction | Part-conditioned pose estimation · IKEA-Manual | Rotation error: 82.79 -> 64.81 | Improved |
| Optimize pose from masks using differentiable rendering instead of learned pose prediction | Part-conditioned pose estimation · IKEA-Manual | CA: 49.53 -> 51.4 | Improved |
| Use differentiable-rendering optimization at inference | SoftRas pose estimation · IKEA-Manual | Inference time (s): <10^{-5} -> 243 | Hurt |
| Evaluate a PartNet-trained assembly model on realistic IKEA-Manual parts rather than PartNet parts | Pretrained part-assembly models · Chair part assembly | Part Accuracy: RGL-Net*: 49.06 -> 3.99; Huang et al.: 39.96 -> 6.90 | Hurt |
| Evaluate a PartNet-trained assembly model on realistic IKEA-Manual parts rather than PartNet parts | Pretrained part-assembly models · Chair part assembly | Shape CD: RGL-Net*: 0.0087 -> 0.0508; Huang et al.: 0.0091 -> 0.0151 | Hurt |
Closest papers by meaning
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.95)
- Two by Two: Learning Multi-Task Pairwise Objects Assembly for Generalizable Robot Manipulation (CVPR 2025 · similarity 0.86)
- Imagine: Image-Guided 3D Part Assembly with Structure Knowledge Graph (AAAI 2025 · similarity 0.85)
- Neural Assembler: Learning to Generate Fine-Grained Robotic Assembly Instructions from Multi-View Images (AAAI 2025 · similarity 0.84)
- Category-Level Multi-Part Multi-Joint 3D Shape Assembly (CVPR 2024 · similarity 0.83)
- Generative 3D Part Assembly via Part-Whole-Hierarchy Message Passing (CVPR 2024 · similarity 0.82)
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.
