AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Convolution with even-sized kernels and symmetric padding

Shuang Wu, Guanrui Wang, Pei Tang, Feng Chen, Luping Shi

NeurIPS 2019 · Architectures and efficiency › CNN design · approach: Large and dynamic kernel design · 50 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use symmetric padding with 2×2 kernels instead of asymmetric padding (ablation)ResNet-50 0.5× · ImageNettop-1 error (%): 31.0 -> 28.4Improved
Use 2×2 symmetric-padding kernels instead of 3×3 kernelsResNet-50 · ImageNettop-1 error (%): 23.8 -> 24.0Hurt
Use 2×2 symmetric-padding kernels instead of 3×3 kernelsDenseNet-121 · ImageNettop-1 error (%): 24.6 -> 25.1Hurt
Use C2sp blocks instead of ShiftNet blocksCompact CNNs · CIFAR100, 56- and 110-layer modelserror (with mixup, %): 21.31 -> 20.91; 20.47 -> 19.52Mixed
Use C2sp blocks instead of depthwise-separable convolution blocksCompact CNNs · CIFAR100memory (MB): 1707 -> 1219; 3170 -> 1951Mixed
Add cutout and mixup augmentationWide-DenseNet C2sp · CIFAR10 testerror (%): 3.54 -> 2.44Improved
Replace 3×3 kernels with C2spWide-DenseNet · CIFAR10 testerror (%): 3.81 -> 3.54Improved
Use C2sp instead of asymmetric C2 kernels (ablation)GAN generator · CIFAR10 generationFID: non-convergence -> 23.35$\pm$0.26Improved
Use C4sp instead of asymmetric C4 kernels (ablation)GAN generator · LSUN bedroom generationFID: 39.17$\pm$5.52 -> 24.61$\pm$2.45Improved
Use C4sp instead of asymmetric C4 kernels (ablation)GAN generator · CelebA-HQ generationFID: 37.93$\pm$7.54 -> 30.22$\pm$3.02Improved
Replace C3 kernels with C2spGAN generator · CIFAR10 generationFID: 24.12$\pm$0.47 -> 23.35$\pm$0.26Improved
Replace C3 kernels with C4spGAN generator · LSUN bedroom and CelebA-HQ generationFID: 36.04$\pm$7.10 -> 24.61$\pm$2.45; 43.39$\pm$5.78 -> 30.22$\pm$3.02Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy