Research map / 3D vision / Face and shape
End-to-End CAD Model Retrieval and 9DoF Alignment in 3D Scans
Armen Avetisyan, Angela Dai, Matthias Niessner
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 differentiable Procrustes alignment loss (ablation) | end-to-end CAD alignment · Scan2CAD val | alignment accuracy (%): 35.74 -> 50.72 | Improved |
| add Procrustes loss versus direct 9DoF l2 regression (ablation) | end-to-end CAD alignment · Scan2CAD val | alignment accuracy (%): 15.12 -> 50.72 | Improved |
| add symmetry-aware dense correspondence (CORRS) prediction (ablation) | end-to-end CAD alignment · Scan2CAD val | alignment accuracy (%): 29.97 -> 50.72 | Improved |
| add explicit symmetry classification (ablation) | CORRS prediction · Scan2CAD val | alignment accuracy (%): 40.51 -> 50.72 | Improved |
| add anchor-style bounding box regression for object detection (ablation) | end-to-end CAD alignment · Scan2CAD val | alignment accuracy (%): 47.64 -> 50.72 | Mixed |
| add anchors (bounding boxes), class-average metric (ablation) | end-to-end CAD alignment · Scan2CAD val | class avg alignment accuracy (%): 42.37 -> 44.61 | Improved |
| add Procrustes loss (ablation) | end-to-end CAD alignment · SUNCG (synthetic scans) | alignment accuracy (%): 36.42 -> 41.83 | Improved |
| end-to-end learned alignment versus decoupled correspondence + RANSAC | Ours vs Scan2CAD · Scan2CAD val | alignment accuracy (%): 31.68 -> 50.72 | Improved |
| end-to-end single forward pass versus Scan2CAD optimization | Ours vs Scan2CAD · runtime, large scene | runtime (s): 740.34s -> 2.60s | Improved |
| end-to-end method on hidden benchmark test set | Ours vs Scan2CAD · Scan2CAD hidden test set | alignment accuracy (%): 28.8 -> 44.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- 3DMatch — 3DMatch: Learning Local Geometric Descriptors From RGB-D Reconstructions (CVPR 2017) · better on 3 of 3 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
- ROCA: Robust CAD Model Retrieval and Alignment From a Single Image (CVPR 2022 · similarity 0.92)
- Scan2CAD: Learning CAD Model Alignment in RGB-D Scans (CVPR 2019 · similarity 0.91)
- FastCAD: Real-Time CAD Retrieval and Alignment from Scans and Videos (ECCV 2024 · similarity 0.90)
- CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention (CVPR 2024 · similarity 0.85)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.85)
- cadrille: Multi-modal CAD Reconstruction with Reinforcement Learning (ICLR 2026 · 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.
