Research map / 3D vision / Face and shape
Learning Canonical View Representation for 3D Shape Recognition With Arbitrary Views
Xin Wei, Yifei Gong, Fudong Wang, Xing Sun, Jian Sun
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 image-level feature encoder (CNN+Transformer) instead of CNN with max-pooling (ablation) | MVCNN-style baseline · ModelNet40 (arbitrary views) | per-instance accuracy (%): 83.20 -> 85.12 | Improved |
| add image-level feature encoder (CNN+Transformer) (ablation) | MVCNN-style baseline · ModelNet40 (arbitrary views) | per-class accuracy (%): 78.86 -> 81.78 | Improved |
| add canonical view representation module (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 85.04 -> 86.91 | Improved |
| use optimal transport rather than Transformer decoder in CVR (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 83.29 -> 86.91 | Improved |
| use optimal transport rather than Transformer decoder in CVR (ablation) | Full model · ModelNet40 (arbitrary views) | per-class accuracy (%): 79.97 -> 84.01 | Improved |
| add canonical view feature separation loss (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 85.32 -> 86.91 | Improved |
| add canonical view feature separation loss (ablation) | Full model · ModelNet40 (arbitrary views) | per-class accuracy (%): 82.28 -> 84.01 | Improved |
| add canonical view aggregator (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 85.75 -> 86.91 | Improved |
| add canonical view aggregator (ablation) | Full model · ModelNet40 (arbitrary views) | per-class accuracy (%): 82.04 -> 84.01 | Improved |
| use number of reference views M=8 rather than 4 (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 85.69 -> 86.91 | Improved |
| use number of reference views M=8 rather than 16 (ablation) | Full model · ModelNet40 (arbitrary views) | per-instance accuracy (%): 86.22 -> 86.91 | Improved |
| use canonical view representation instead of MVCNN-M/GVCNN-M under object rotation | Proposed vs MVCNN-M · ModelNet40 (rotated) | per-instance accuracy (%): 88.17 -> 92.22 | 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.
- View-GCN — View-GCN: View-Based Graph Convolutional Network for 3D Shape Analysis (CVPR 2020) · 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
- Learning View-Based Graph Convolutional Network for Multi-View 3D Shape Analysis (TPAMI 2022 · similarity 0.88)
- Learning Canonical 3D Object Representation for Fine-Grained Recognition (ICCV 2021 · similarity 0.87)
- Beyond Viewpoint: Robust 3D Object Recognition under Arbitrary Views through Joint Multi-Part Representation (ECCV 2024 · similarity 0.87)
- Pixels, Voxels, and Views: A Study of Shape Representations for Single View 3D Object Shape Prediction (CVPR 2018 · similarity 0.87)
- View-GCN: View-Based Graph Convolutional Network for 3D Shape Analysis (CVPR 2020 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · 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.
