Research map / Architectures and efficiency / CNN design
Wavelet Integrated CNNs for Noise-Robust Image Classification
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 down-sampling with DWT low-frequency outputs | ResNet18 · ImageNet validation | Top-1 accuracy: 69.76 -> 71.62 | Helped |
| Replace standard down-sampling with DWT low-frequency outputs | ResNet18 · ImageNet-C noise | Noise mCE: 88.97 -> 80.38 | Helped |
| Replace SegNet pooling and unpooling with DWT and IDWT | SegNet · CamVid test | mIoU: 57.89 -> 64.23 | Helped |
| Use Cohen ch3.3 rather than the baseline down-sampling | VGG16bn · ImageNet validation | Top-1 accuracy: 73.37 -> 74.40 | Helped |
| Use symmetric Haar or Cohen wavelets rather than baseline down-sampling | VGG16bn · ImageNet-C noise | Noise mCE | No clear effect |
| Use Daubechies wavelets with longer filters (ablation) | ResNet18 · ImageNet validation | Top-1 accuracy: 71.48 -> 68.74 | Hurt |
| Use Daubechies wavelets with longer filters (ablation) | ResNet18 · ImageNet-C noise | Noise mCE: 88.97 -> 83.77 | Helped |
| Average multiple wavelet components rather than retain only the low-frequency component (ablation) | ResNet18 · ImageNet | Top-1 accuracy: 71.47 -> 70.06 | Hurt |
| Average multiple wavelet components rather than retain only the low-frequency component (ablation) | ResNet18 · ImageNet-C noise | Noise mCE: 80.91 -> 86.02 | Hurt |
| Concatenate all DWT components rather than retain only the low-frequency component (ablation) | ResNet18 · ImageNet | Top-1 accuracy: +0.11% to +0.77% | Helped |
| Concatenate all DWT components rather than retain only the low-frequency component (ablation) | ResNet18 · ImageNet-C noise | Noise mCE | No clear effect |
Closest papers by meaning
- Wavelet Pooling for Convolutional Neural Networks (ICLR 2018 · similarity 0.86)
- Learning a Wavelet-Like Auto-Encoder to Accelerate Deep Neural Networks (AAAI 2018 · similarity 0.85)
- Wavelet Feature Maps Compression for Image-to-Image CNNs (NeurIPS 2022 · similarity 0.85)
- Task-Driven Wavelets using Constrained Empirical Risk Minimization (CVPR 2024 · similarity 0.84)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · similarity 0.82)
- Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional Neural Networks (AAAI 2024 · similarity 0.82)
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.