Research map / Architectures and efficiency / CNN design
ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks
Xiaohan Ding, Yuchen Guo, Guiguang Ding, Jungong Han
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 3x3 layers with ACB (add 1x3 and 3x1 branches) | ResNet-56 · CIFAR-10 | top-1 accuracy (%): 94.31 -> 95.09 | Improved |
| replace 3x3 layers with ACB | Cifar-quick · CIFAR-10 | top-1 accuracy (%): 83.13 -> 84.24 | Improved |
| replace 3x3 layers with ACB | DenseNet-40 · CIFAR-100 | top-1 accuracy (%): 73.14 -> 73.41 | Improved |
| replace 3x3 layers with ACB | WRN-16-8 · CIFAR-100 | top-1 accuracy (%): 78.65 -> 79.44 | Improved |
| replace 3x3 layers with ACB | AlexNet · ImageNet | top-1 accuracy (%): 55.92 -> 57.44 | Improved |
| replace 3x3 layers with ACB | ResNet-18 · ImageNet | top-1 accuracy (%): 70.36 -> 71.14 | Improved |
| replace 3x3 layers with ACB | DenseNet-121 · ImageNet | top-1 accuracy (%): 75.15 -> 75.82 | Improved |
| add horizontal kernel to ACB (ablation) | ACNet AlexNet · ImageNet | top-1 accuracy (%): 57.10 -> 57.44 | Improved |
| add vertical kernel to ACB (ablation) | ACNet AlexNet · ImageNet | top-1 accuracy (%): 57.25 -> 57.44 | Improved |
| add batch norm in each branch (ablation) | ACNet AlexNet · ImageNet | top-1 accuracy (%): 56.18 -> 57.44 | Improved |
| add horizontal kernel to ACB (ablation) | ACNet ResNet-18 · ImageNet | top-1 accuracy (%): 70.78 -> 71.14 | Improved |
| add vertical kernel to ACB (ablation) | ACNet ResNet-18 · ImageNet | top-1 accuracy (%): 70.70 -> 71.14 | Improved |
Closest papers by meaning
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 2024 · similarity 0.89)
- Asymmetric Convolution: An Efficient and Generalized Method to Fuse Feature Maps in Multiple Vision Tasks (TPAMI 2024 · similarity 0.86)
- The Unreasonable Effectiveness of Fully-Connected Layers for Low-Data Regimes (NeurIPS 2022 · similarity 0.85)
- Refining Architectures of Deep Convolutional Neural Networks (CVPR 2016 · similarity 0.85)
- VideoOneNet: Bidirectional Convolutional Recurrent OneNet with Trainable Data Steps for Video Processing (ICML 2020 · similarity 0.85)
- Ubernet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory (CVPR 2017 · similarity 0.85)
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.
