AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Dynamic Group Convolution for Accelerating Convolutional Neural Networks

Zhuo Su, Linpu Fang, Wenxiong Kang, Dewen Hu, Matti Pietikäinen, Li Liu

ECCV 2020 · Architectures and efficiency › CNN design · approach: Efficient CNN scaling and dynamic inference · 64 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 dynamic per-sample channel selection (DGC) instead of fixed learned group conv (FLGC)MobileNetV2 · CIFAR-10top-1 error (%): 5.89 -> 5.17Improved
use DGC instead of learned group convolution (LGC)CondenseNet-50 · CIFAR-10top-1 error (%): 6.10 -> 5.83Improved
use DGC instead of LGCCondenseNet-50 · CIFAR-100top-1 error (%): 26.58 -> 26.18Improved
use DGC instead of standard group conv (SGC)CondenseNet · ImageNettop-1 error (%): 29.0 -> 25.4Improved
use DGC instead of static filter pruning FPGMResNet-18 · ImageNettop-1 error (%): 31.59 -> 31.22Improved
use DGC instead of dynamic channel pruning FBSResNet-18 · ImageNettop-1 error (%): 31.83 -> 31.22Improved
use global threshold instead of head-wise threshold (DGC-G)ResNet-18 · ImageNettop-1 error (%): 31.22 -> 31.37Hurt
use global threshold instead of head-wise threshold (DGC-G)CondenseNet · ImageNettop-1 error (%): 25.4 -> 25.2Improved
use global threshold instead of head-wise threshold (DGC-G)CondenseNet-86 · CIFAR-10top-1 error (%): 4.77 -> 4.42Improved
add angle enlargement loss with global threshold (ablation)CondenseNet-DGC-G · ImageNettop-1 error (%): 25.8 -> 25.2Improved
add angle enlargement loss with head-wise threshold (ablation)DGC · ImageNettop-1 error (%)No clear effect
increase number of heads from 1 to 4, then 8 (ablation)CondenseNet-DGC-50 · CIFAR-10top-1 error (%)Mixed

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