Research map / Architectures and efficiency / CNN design
Frequency Domain Compact 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 |
|---|---|---|---|
| Use learned temporal-frequency pruning and channel pruning instead of regularization-based pruning | 3D-ResNet-18 · UCF101, 2× speed-up | increased error (%): 0.41 -> 0.10 | Helped |
| Use learned temporal-frequency pruning and channel pruning instead of regularization-based pruning | 3D-ResNet-18 · UCF101, 4× speed-up | increased error (%): 2.87 -> 2.16 | Helped |
| Apply frequency-domain temporal pruning in addition to channel pruning (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 2× speed-up | increased error (%): 0.81 -> 0.10 | Helped |
| Apply frequency-domain temporal pruning in addition to channel pruning (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 4× speed-up | increased error (%): 3.58 -> 2.16 | Helped |
| Learn an optimal transform instead of using fixed DCT (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 2× speed-up | increased error (%): 0.42 -> 0.10 | Helped |
| Learn an optimal transform instead of using fixed DCT (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 4× speed-up | increased error (%): 2.96 -> 2.16 | Helped |
| Replace learned transform with fixed DCT for temporal-frequency pruning (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 2× speed-up | increased error (%): 0.10 -> 0.42 | Hurt |
| Replace learned transform with fixed DCT for temporal-frequency pruning (ablation) | FDC3D on 3D-ResNet-18 · UCF101, 4× speed-up | increased error (%): 2.16 -> 2.96 | Hurt |
| Apply 2× FDC3D compression | Residual 3D U-Net · BraTS 2018 | Dice (ET): 0.7974 -> 0.7832 | Hurt |
| Apply 2× FDC3D compression | Residual 3D U-Net · BraTS 2018 | Dice (WT): 0.7971 -> 0.7831 | Hurt |
| Apply 2× FDC3D compression | Residual 3D U-Net · BraTS 2018 | Dice (TC): 0.6908 -> 0.6816 | Hurt |
Closest papers by meaning
- CNNpack: Packing Convolutional Neural Networks in the Frequency Domain (NeurIPS 2016 · similarity 0.88)
- Frequency-Domain Dynamic Pruning for Convolutional Neural Networks (NeurIPS 2018 · similarity 0.88)
- LP-3DCNN: Unveiling Local Phase in 3D Convolutional Neural Networks (CVPR 2019 · similarity 0.86)
- Learning in the Frequency Domain (CVPR 2020 · similarity 0.86)
- Frequency Dynamic Convolution for Dense Image Prediction (CVPR 2025 · similarity 0.84)
- Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional Neural Networks (AAAI 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.