AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Beyond Self-Attention: External Attention Using Two Linear Layers for Visual Tasks

Meng-Hao Guo, Zheng-Ning Liu, Tai‐Jiang Mu, Shi‐Min Hu

TPAMI 2022 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 728 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
replace self-attention with external attention (DoubleNorm) (ablation)FCN ResNet-50 · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 76.6 -> 77.4Improved
add double normalization instead of softmax in external attention (ablation)FCN + EA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 75.3 -> 77.4Improved
use DoubleNorm instead of softmax in self-attention (ablation)FCN + SA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 76.2 -> 76.6Improved
increase memory units from 8 to 64 (ablation)FCN + EA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 77.1 -> 77.4Improved
increase memory units from 64 to 256 (ablation)FCN + EA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 77.4 -> 77.0Hurt
reduce output stride from 16 to 8 (ablation)FCN + EA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 77.4 -> 77.8Improved
use ResNet-101 instead of ResNet-50 (ablation)FCN + EA · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 77.4 -> 78.3Improved
add external attention to FCN (ablation)FCN ResNet-50 · PASCAL VOC val (FCN ablation, 30k iters)mIoU: 75.7 -> 77.4Improved
use multi-head external attention instead of single-head (ablation)T2T-ViT-7 Performer MEA · ImageNet-1Ktop-1 accuracy: 63.2 -> 68.6Improved
increase input size from 224 to 384 (ablation)T2T-ViT-7 MEA · ImageNet-1Ktop-1 accuracy: 68.6 -> 70.9Improved
replace LN by BN in MLP blocks (ablation)EAMLP-7 · ImageNet-1Ktop-1 accuracy: 68.9 -> 70.0Improved
replace LN by BN in MLP blocks (ablation)EAMLP-14/19 · ImageNet-1Ktop-1 accuracy: 78.9 -> FailedHurt

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