Research map / Architectures and efficiency / CNN design
LP-3DCNN: Unveiling Local Phase in 3D Convolutional Neural Networks
Check 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 standard 3D convolutions with ReLPV blocks | VoxNet, VoxNetPlus, and LightNet · ModelNet10 and ModelNet40 | accuracy: 83 -> 86.26; 92 -> 92.24; 83.91 -> 88.1; 93.36 -> 93.4; 86.90 -> 87.5; 93.39 -> 92.95 | Helped |
| Use the proposed ReLPV network with Az×24 augmentation instead of Az×12 | LP-3DCNN · ModelNet40 and ModelNet10 | accuracy: 89.4 -> 92.1; 93.76 -> 94.4 | Helped |
| Replace standard 3D convolutions with ReLPV blocks | mC3D · UCF-101 split-1 | accuracy: 44 -> 53.58; 42.5 -> 51.44; 42.3 -> 50.54; 36.17 -> 48.99 | Helped |
| Increase the ReLPV local volume from 3 to 9 (ablation) | LP-mC3D · UCF-101 split-1 | accuracy: 53.58 -> 48.99 | Hurt |
| Increase output feature maps from 64 to 256 (ablation) | LP-mC3D with local volume 3 · UCF-101 split-1 | accuracy: 50.96 -> 53.50 | Helped |
| Increase output feature maps from 64 to 256 (ablation) | LP-mC3D with local volume 5 · UCF-101 split-1 | accuracy: 50.29 -> 53.22 | Helped |
| Increase output feature maps from 64 to 256 (ablation) | LP-mC3D with local volume 7 · UCF-101 split-1 | accuracy: 47.66 -> 51.14 | Helped |
| Replace top convolutional layers with ReLPV blocks (ablation) | mC3D₃ hybrid, one top layer replaced · UCF-101 split-1 | accuracy: 44 -> 51.51 | Helped |
| Replace three bottom convolutional layers with ReLPV blocks (ablation) | mC3D₃ hybrid · UCF-101 split-1 | accuracy: 44 -> 40.84 | Hurt |
| Replace all convolutional layers with ReLPV blocks rather than using a hybrid (ablation) | mC3D₃ · UCF-101 split-1 | accuracy: 51.51 -> 53.58 | Helped |
Closest papers by meaning
- Frequency Domain Compact 3D Convolutional Neural Networks (CVPR 2020 · similarity 0.86)
- T-C3D: Temporal Convolutional 3D Network for Real-Time Action Recognition (AAAI 2018 · similarity 0.86)
- Volumetric and Multi-View CNNs for Object Classification on 3D Data (CVPR 2016 · similarity 0.85)
- Efficient 3D Recognition with Event-driven Spike Sparse Convolution (AAAI 2025 · similarity 0.84)
- Rethinking 3D Convolution in $\ell_p$-norm Space (NeurIPS 2024 · similarity 0.84)
- OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmentation (CVPR 2024 · similarity 0.84)
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.