Research map / 3D vision / Face and shape
GVCNN: Group-View Convolutional Neural Networks for 3D Shape Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 direct view-to-shape pooling with hierarchical view grouping and weighted group fusion | GoogLeNet multi-view CNN · ModelNet40 | accuracy: 92.0% -> 93.1% | Helped |
| Replace direct view-to-shape pooling with hierarchical view grouping and weighted group fusion | GoogLeNet multi-view CNN · ModelNet40 | mAP: 74.62% -> 79.7% | Helped |
| Replace direct view-to-shape pooling with hierarchical view grouping and weighted group fusion | GoogLeNet multi-view CNN · ModelNet40 | accuracy: 92.2% -> 92.6% | Helped |
| Replace direct view-to-shape pooling with hierarchical view grouping and weighted group fusion | GoogLeNet multi-view CNN · ModelNet40 | mAP: 74.1% -> 81.3% | Helped |
| Add low-rank Mahalanobis metric learning | GVCNN, 8 views · ModelNet40 | mAP: 79.7% -> 84.5% | Helped |
| Add low-rank Mahalanobis metric learning | GVCNN, 12 views · ModelNet40 | mAP: 81.3% -> 85.7% | Helped |
| Add low-rank Mahalanobis metric learning | GVCNN, 8 views · ModelNet40 | accuracy: 93.1% -> 93.1% | No clear effect |
| Add low-rank Mahalanobis metric learning | GVCNN, 12 views · ModelNet40 | accuracy: 92.6% -> 92.6% | No clear effect |
| Use fewer test views than the 8-view training configuration (ablation) | GVCNN trained with 8 views · ModelNet40 | accuracy: 93.1% -> 70.0%; 93.1% -> 71.2% | Hurt |
| Use fewer test views than the 12-view training configuration (ablation) | GVCNN trained with 12 views · ModelNet40 | accuracy: 92.6% -> 75.0%; 92.6% -> 76.8% | Hurt |
| Test the 8-view-trained model using 12 views (ablation) | GVCNN trained with 8 views · ModelNet40 | accuracy: 93.1% -> 91.5% | Hurt |
| Test the 12-view-trained model using 8 views (ablation) | GVCNN trained with 12 views · ModelNet40 | accuracy: 92.6% -> 92.1% | Hurt |
Closest papers by meaning
- View-GCN: View-Based Graph Convolutional Network for 3D Shape Analysis (CVPR 2020 · similarity 0.93)
- MLVCNN: Multi-Loop-View Convolutional Neural Network for 3D Shape Retrieval (AAAI 2019 · similarity 0.90)
- Group-Pair Convolutional Neural Networks for Multi-View Based 3D Object Retrieval (AAAI 2018 · similarity 0.90)
- FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis (CVPR 2018 · similarity 0.87)
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · 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.