Research map / Architectures and efficiency / CNN design
Doubly Convolutional Neural Networks
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 standard convolutional layers with doubly convolutional layers | CNN · CIFAR-10, without data augmentation | test error: 9.85% -> 8.58% | Helped |
| Replace standard convolutional layers with doubly convolutional layers | CNN · CIFAR-100, with data augmentation | test error: 33.04% -> 26.53% | Helped |
| Replace standard convolutional layers with doubly convolutional layers | CNN · ImageNet validation | Top-1 error: 29.42% -> 26.27 % | Helped |
| Add doubly convolutional layers at layers 1-2 (ablation) | CNN · CIFAR-10 | test error: 9.85% -> 9.12% | Helped |
| Add doubly convolutional layers at layers 3-4 (ablation) | CNN · CIFAR-10 | test error: 9.85% -> 9.23% | Helped |
| Add doubly convolutional layers at layers 5-6 (ablation) | CNN · CIFAR-10 | test error: 9.85% -> 9.45% | Helped |
| Add doubly convolutional layers at layers 7-8 (ablation) | CNN · CIFAR-10 | test error: 9.85% -> 9.57% | Helped |
| Use DCNN configuration 32-6-3-2 with the same parameter count as CNN (ablation) | DCNN · CIFAR-10 | test error: 9.85% -> 9.05% | Helped |
| Use DCNN configuration 4-10-3-1 with fewer parameters (ablation) | DCNN · CIFAR-100 | test error: 34.26% -> 33.57% | Helped |
Closest papers by meaning
- CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters (CVPR 2022 · similarity 0.88)
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.88)
- More Is Less: A More Complicated Network With Less Inference Complexity (CVPR 2017 · similarity 0.88)
- DCFNet: Deep Neural Network with Decomposed Convolutional Filters (ICML 2018 · similarity 0.88)
- Modulated Convolutional Networks (CVPR 2018 · similarity 0.88)
- Learning Versatile Filters for Efficient Convolutional Neural Networks (NeurIPS 2018 · similarity 0.87)
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.