Research map / Architectures and efficiency / CNN design
Densely Connected Convolutional 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 |
|---|---|---|---|
| Increase growth rate from k=12 to k=24 (ablation) | DenseNet (L=100) · CIFAR-10 | error rate (%): 5.77 to 5.83 | Hurt |
| Increase growth rate from k=12 to k=24 (ablation) | DenseNet (L=100) · CIFAR-10+ | error rate (%): 4.10 to 3.74 | Helped |
| Increase growth rate from k=12 to k=24 (ablation) | DenseNet (L=100) · CIFAR-100+ | error rate (%): 20.20 to 19.25 | Helped |
| Add bottleneck layers and transition-layer compression (ablation) | DenseNet · CIFAR-10+ | error rate (%): 5.24 -> 4.51 | Helped |
| Add bottleneck layers and transition-layer compression (ablation) | DenseNet · CIFAR-100+ | error rate (%): 20.20 -> 22.27 | Hurt |
| Increase DenseNet-BC depth from 100 to 250 layers (ablation) | DenseNet-BC (k=24) · SVHN | error rate (%): 1.74 -> 1.74 | Hurt |
| Increase DenseNet-BC depth from 100 to 250 layers (ablation) | DenseNet-BC (k=24) · CIFAR-10+ | error rate (%): 4.51 -> 3.62 | Helped |
| Increase DenseNet-BC depth from 100 to 250 layers (ablation) | DenseNet-BC (k=24) · CIFAR-100+ | error rate (%): 22.27 -> 17.60 | Helped |
| Use dense concatenation connections instead of conventional layer-by-layer connections | DenseNet · CIFAR-10 and CIFAR-100 | error rate (%) | Helped |
| Use dense connections to reuse earlier feature maps | DenseNet · ImageNet validation | parameters at comparable error: 20M; more than 40M | Helped |
Closest papers by meaning
- Contextual Convolutional Networks (ICLR 2023 · similarity 0.89)
- Densely Connected Multi-Dilated Convolutional Networks for Dense Prediction Tasks (CVPR 2021 · similarity 0.89)
- Multi-Scale Dense Networks for Resource Efficient Image Classification (ICLR 2018 · similarity 0.89)
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.88)
- Deep Residual Learning for Image Recognition (CVPR 2016 · similarity 0.88)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · 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.