Research map / 3D vision / 3D generation
DeepCCFV: Camera Constraint-Free Multi-View Convolutional Neural Network for 3D Object Retrieval
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 |
|---|---|---|---|
| add DropMax blocks (DeepCCFV) to MVCNN | MVCNN VGG11-BN · ModelNet40 retrieval, 1 view | mAP: 61.91% -> 79.57% | Helped |
| add DropMax blocks (DeepCCFV) to MVCNN | MVCNN VGG11-BN · ModelNet40 retrieval, 12 views | mAP: 85.52% -> 86.92% | Helped |
| add DropMax blocks (DeepCCFV) to MVCNN | MVCNN ResNet50 · ModelNet40 retrieval, 1 view | mAP: 83.25% -> 35.75% higher than basic MVCNN | Helped |
| add DropMax (DeepCCFV) under random camera positions | MVCNN VGG11-BN · ModelNet40 retrieval random cameras, 12 views | mAP: 75.96% -> 82.87% | Helped |
| add Dropout to MVCNN (ablation) | MVCNN VGG11-BN · ModelNet40 retrieval, random cameras | mAP: 75.96% -> 74.86% | Hurt |
| add Dropview to MVCNN (ablation) | MVCNN ResNet50 · ModelNet40 retrieval | mAP: drop between extremes 38.14% -> 7.92% | Helped |
| add Dropview to MVCNN (ablation) | MVCNN VGG11-BN · ModelNet40 retrieval, predefined cameras | mAP: 23.61% -> 6.89% gap between worst and best | Mixed |
| add DropMax (DeepCCFV) for cross-modal retrieval with DGCNN | MVCNN+DGCNN · ModelNet40 cross-modal, 1view to point cloud | mAP: 36.88% -> 75.20% | Helped |
| add DropMax (DeepCCFV) for classification | MVCNN VGG11-BN · ModelNet40 classification, 1 view | accuracy: 64.28% -> 82.11% | Helped |
| add DropMax (DeepCCFV) for classification at full views | MVCNN VGG11-BN · ModelNet40 classification, 12 views | accuracy: 92.75% -> 92.30% | No clear effect |
| vary DropMax probability p from 0.1 to 0.8 (ablation) | DeepCCFV · ModelNet40 retrieval | mAP | No clear effect |
| vary DropMax k1 from 300 to 600 and k2 from 0 to 400 (ablation) | DeepCCFV · ModelNet40 retrieval | mAP | No clear effect |
Closest papers by meaning
- Group-Pair Convolutional Neural Networks for Multi-View Based 3D Object Retrieval (AAAI 2018 · similarity 0.88)
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · similarity 0.86)
- What Do Single-View 3D Reconstruction Networks Learn? (CVPR 2019 · similarity 0.86)
- Multi-View Attentive Contextualization for Multi-View 3D Object Detection (CVPR 2024 · similarity 0.84)
- Duoduo CLIP: Efficient 3D Understanding with Multi-View Images (ICLR 2025 · similarity 0.83)
- DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision (CVPR 2024 · similarity 0.83)
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.