Research map / 3D vision / Face and shape
Mask2CAD: 3D Shape Prediction by Learning to Segment and Retrieve
Weicheng Kuo, Anelia Angelova, Tsung-Yi Lin, Angela Dai
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 all CAD models at test time without retraining (ablation) | Mask2CAD · Pix3D S2 | AP mesh: 6.5 -> 8.2 | Improved |
| add all CAD models at test time without retraining (ablation) | Mask2CAD · Pix3D S2 | AP50 mesh: 17.3 -> 20.7 | Improved |
| add all CAD models at test time without retraining (ablation) | Mask2CAD · Pix3D S2 | AP75 mesh: 3.8 -> 4.8 | Improved |
| replace predicted shape with ground truth shape (ablation) | Mask2CAD · Pix3D S1 | AP mesh: +14.6 | Improved |
| replace predicted rotation with ground truth (ablation) | Mask2CAD · Pix3D S1 | AP mesh: +10.2 | Improved |
| replace predicted translation with ground truth (ablation) | Mask2CAD · Pix3D S1 | AP mesh: +7.5 | Improved |
| replace predicted 2D boxes with ground truth boxes (ablation) | Mask2CAD · Pix3D S1 | AP mesh | No clear effect |
| retrieval-based Mask2CAD instead of Mesh R-CNN | Mask2CAD · Pix3D S1 | AP mesh: 17.2 -> 33.2 | Improved |
| retrieval-based Mask2CAD instead of Mesh R-CNN | Mask2CAD · Pix3D S1 | AP75 mesh: 7.4 -> 30.8 | Improved |
| Mask2CAD versus Mesh R-CNN on tool category | Mask2CAD · Pix3D S1 | AP mesh (tool): 6.5 -> 4.2 | Hurt |
| Mask2CAD versus Mesh R-CNN with GT boxes | Mask2CAD · Pix3D S1 (GT boxes) | F1@0.3: 56.4 -> 66.4 | Improved |
| Mask2CAD versus FroDO trained with ShapeNet | Mask2CAD · Pix3D S1 chairs | IoU: 0.325 -> 0.613 | Improved |
Later papers that compared with this method
1 later accepted paper reports Mask2CAD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Patch2CAD: Patchwise Embedding Learning for In-the-Wild Shape Retrieval From a Single Image (ICCV 2021 · better on 7 of 10 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
- Patch2CAD: Patchwise Embedding Learning for In-the-Wild Shape Retrieval From a Single Image (ICCV 2021 · similarity 0.90)
- FastCAD: Real-Time CAD Retrieval and Alignment from Scans and Videos (ECCV 2024 · similarity 0.90)
- Scan2CAD: Learning CAD Model Alignment in RGB-D Scans (CVPR 2019 · similarity 0.88)
- Straight to Shapes: Real-Time Detection of Encoded Shapes (CVPR 2017 · similarity 0.88)
- Mesh R-CNN (ICCV 2019 · similarity 0.88)
- Mask3D: Pre-Training 2D Vision Transformers by Learning Masked 3D Priors (CVPR 2023 · 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.
