AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Contextual Transformer Networks for Visual Recognition

Yehao Li, Ting Yao, Yingwei Pan, Tao Mei

TPAMI 2022 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 795 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 dynamic self-attention context to static context (ablation)CoTNet-50 · ImageNet-1k val (default training, CoTNet-50)top-1 accuracy: 77.1 -> 78.5Improved
fuse static and dynamic contexts by linear summation (ablation)CoTNet-50 · ImageNet-1k val (default training, CoTNet-50)top-1 accuracy: 78.5 -> 78.7Improved
fuse static and dynamic contexts via attention instead of summation (ablation)CoTNet-50 · ImageNet-1k val (default training, CoTNet-50)top-1 accuracy: 78.7 -> 79.2Improved
replace 3x3 convs in res4 as well as res5 (ablation)CoTNet-50 · ImageNet-1k val (default training, CoTNet-50)top-1 accuracy: 78.5 -> 79.0Improved
replace 3x3 convs in res2 and res3 as well (ablation)CoTNet-50 · ImageNet-1k val (default training, CoTNet-50)top-1 accuracy: 79.0 -> 79.2Improved
replace 3x3 convs with CoT blocks across all stagesResNet-50 · ImageNet-1k val (default training)top-1 accuracy: 77.3 -> 79.2Improved
use CoT in res4 and res5 on SE-ResNetD-50 (ablation)SE-ResNetD-50 · ImageNet-1k val (default training)top-1 accuracy: 79.1 -> 79.8Improved
add exponential moving averageCoTNet-50 · ImageNet-1k val (default training)top-1 accuracy: 79.2 -> 79.8Improved
use advanced training (350 epochs, RandAugment, mixup, dropout)CoTNet-50 · ImageNet-1k valtop-1 accuracy: 79.2 -> 81.3Improved
use ResNeXt-style CoTNeXt instead of CoTNetCoTNet-50 · ImageNet-1k val (default training)top-1 accuracy: 79.2 -> 79.5Improved
use CoTNet backbone instead of ResNet backboneFaster-RCNN · COCO val Faster-RCNNAP: 39.34 -> 43.50Improved
use CoTNeXt-101 backbone instead of ResNeSt-101Cascade-RCNN · COCO val Cascade-RCNNAP: 47.51 -> 49.02Improved

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