Research map / 3D vision / Face and shape
Joint Embedding of 3D Scan and CAD Objects
Manuel Dahnert, Angela Dai, Leonidas J. Guibas, 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 scan segmentation hourglass (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (class avg): 0.19 -> 0.39 | Improved |
| add scan completion hourglass (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (class avg): 0.16 -> 0.39 | Improved |
| add segmentation and completion together (ablation) | joint scan-CAD embedding · Scan-CAD embedding confusion (class avg) | confusion (class avg): 0.16 -> 0.50 | Improved |
| add segmentation (ablation) | joint scan-CAD embedding · Scan-CAD embedding confusion (class avg) | confusion (class avg): 0.49 -> 0.50 | No clear effect |
| add completion (ablation) | joint scan-CAD embedding · Scan-CAD embedding confusion (class avg) | confusion (class avg): 0.24 -> 0.50 | Improved |
| train with negative samples (triplet loss) instead of positives only (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (class avg): 0.10 -> 0.39 | Improved |
| add triplet negatives (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (ranking quality, inst avg) | ranking quality (inst avg): 0.04 -> 0.16 | Improved |
| train end-to-end (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (class avg): 0.29 -> 0.39 | Improved |
| use own completion network instead of 3D-EPN (ablation) | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (class avg): 0.30 -> 0.39 | Improved |
| use own segmentation instead of SGPN | scan segmentation · Scan segmentation/completion IoU | IoU: 0.10 -> 0.36 | Improved |
| use own completion instead of 3D-EPN after segmentation | scan completion · Scan segmentation/completion IoU | IoU: 0.48 -> 0.53 | Improved |
| train with rotation augmentation and test on random rotations | joint scan-CAD embedding · Scan-CAD Object Similarity test (retrieval accuracy, inst avg) | top-1 retrieval accuracy (inst avg): 0.43 -> 0.42 | No clear effect |
Closest papers by meaning
- End-to-End CAD Model Retrieval and 9DoF Alignment in 3D Scans (ICCV 2019 · similarity 0.89)
- CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention (CVPR 2024 · similarity 0.86)
- Point2CAD: Reverse Engineering CAD Models from 3D Point Clouds (CVPR 2024 · similarity 0.85)
- From 2D CAD Drawings to 3D Parametric Models: A Vision-Language Approach (AAAI 2025 · similarity 0.85)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.85)
- KP-RED: Exploiting Semantic Keypoints for Joint 3D Shape Retrieval and Deformation (CVPR 2024 · similarity 0.84)
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.
