Research map / 3D vision / Face and shape
Equivariant Multi-View Networks
Carlos Esteves, Yinshuang Xu, Christine Allen-Blanchette, Kostas Daniilidis
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 |
|---|---|---|---|
| replace view pooling with group convolution aggregation (Ours-12 vs MVCNN-M-12) | multi-view CNN · ModelNet40 (aligned) | mAP: 89.13 -> 91.82 | Improved |
| replace view pooling with group convolution aggregation (Ours-60 vs MVCNN-M-60) | multi-view CNN · ModelNet40 (rotated) | mAP: 78.18 -> 82.61 | Improved |
| replace view pooling with group convolution aggregation (Ours-60 vs MVCNN-M-60) | multi-view CNN · SHREC'17 (rotated) | score: 57.5 -> 62.1 | Improved |
| use polar view representation on homogeneous space (Ours-R-60 vs Ours-60) | equivariant multi-view network · ModelNet40 (rotated) | mAP: 82.61 -> 88.57 | Improved |
| use 60 views instead of 12 on aligned data | equivariant multi-view network · ModelNet40 (aligned) | accuracy: 94.51 -> 94.36 | No clear effect |
| increase views from 12 to 60 on rotated data | equivariant multi-view network · ModelNet40 (rotated) | accuracy: 88.50 -> 91.00 | Improved |
| add filter support of 9 elements instead of 3 (ablation) | Ours-60 · ModelNet40 (rotated) | mAP: 80.49 -> 82.61 | Improved |
| use 3 G-Conv layers instead of 1 (ablation) | Ours-60 · ModelNet40 (rotated) | mAP: 79.59 -> 82.61 | Improved |
| use 3 G-Conv layers instead of 2 (ablation) | Ours-60 · ModelNet40 (rotated) | accuracy: 91.00 -> 91.00 | No clear effect |
| use all 60 views instead of 30 at test time (ablation) | Ours-60 · ModelNet40 (rotated) | mAP: 79.20 -> 82.61 | Improved |
| use all 60 views instead of 5 (ablation) | Ours-60 · ModelNet40 (rotated) | mAP: 64.88 -> 82.61 | Improved |
| use ImageNet pretrained CNN backbone (ablation) | Ours-60 · ModelNet40 (rotated) | mAP: 70.44 -> 82.61 | 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.
- PVNet — PVNet: Pixel-Wise Voting Network for 6DoF Pose Estimation (CVPR 2019) · better on 1 of 1 shared result
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
- $SE(3)$ Equivariant Ray Embeddings for Implicit Multi-View Depth Estimation (NeurIPS 2024 · similarity 0.89)
- General E(2)-Equivariant Steerable CNNs (NeurIPS 2019 · similarity 0.88)
- Attentive Group Equivariant Convolutional Networks (ICML 2020 · similarity 0.88)
- GVCNN: Group-View Convolutional Neural Networks for 3D Shape Recognition (CVPR 2018 · similarity 0.88)
- E2PN: Efficient SE(3)-Equivariant Point Network (CVPR 2023 · similarity 0.88)
- Probing Equivariance and Symmetry Breaking in Convolutional Networks (NeurIPS 2025 · similarity 0.87)
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.
