AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

An Empirical Study of Spatial Attention Mechanisms in Deep Networks

Xizhou Zhu, Dazhi Cheng, Zheng Zhang, Stephen Lin, Jifeng Dai

ICCV 2019 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 679 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
add Transformer attention (full 1111) over no attention (ablation)ResNet-50 detector · COCO object detectionmAP: 36.4 -> 38.8Improved
add Transformer attention (full 1111) over no attention (ablation)Transformer base · WMT14 En-DeBLEU: 20.9 -> 27.7Improved
add query and key content term E1 to self-attention (ablation)Transformer attention · CityscapesmIoU: 76.6 -> 76.7No clear effect
add query and key content term E1 to encoder-decoder attention (ablation)Transformer attention · WMT14 En-DeBLEUImproved
add key content only term E3 to E2 (ablation)Transformer attention · COCO object detectionmAP: 38.6 -> 38.8Improved
replace E2 with deformable convolution (ablation)Transformer attention · COCO object detectionmAP: 38.6 -> 39.9Improved
replace E2 with deformable convolution (ablation)Transformer attention · CityscapesmIoU: 76.3 -> 77.2Improved
replace E2 with deformable convolution (ablation)Transformer attention · WMT14 En-DeBLEU: 27.4 -> 27.3No clear effect
add key content only term E3 to deformable convolution (ablation)deformable convolution · COCO object detectionmAP: 39.9 -> 40.8Improved
replace E2 with dynamic convolution (kernel 31) (ablation)Transformer attention · COCO object detectionmAP: 38.6 -> 37.9Hurt
replace E2 with dynamic convolution (kernel 31) (ablation)Transformer attention · CityscapesmIoU: 75.5 -> 74.2Hurt
replace E2 with dynamic convolution (kernel 31) (ablation)Transformer attention · WMT14 En-DeBLEU: 27.4 -> 27.6No clear effect

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