AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Gaussian Context Transformer

Dongsheng Ruan, Daiyin Wang, Yuan Zheng, Nenggan Zheng, Min Zheng

CVPR 2021 · Architectures and efficiency › CNN design · approach: Efficient CNN scaling and dynamic inference · 50 citations (OpenAlex)

Check 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 parameter-free GCT-B0 channel attentionResNet50 · ImageNet validationTop-1 accuracy: 76.15 -> 77.51Improved
Add learnable GCT-B1 channel attentionResNet50 · ImageNet validationTop-1 accuracy: 76.15 -> 77.55Improved
Use learnable rather than fixed Gaussian standard deviationResNet50 + GCT · ImageNet validationTop-1 accuracy: 77.51 -> 77.55Improved
Add GCT-B0 channel attentionResNet101 · ImageNet validationTop-1 accuracy: 77.37 -> 78.60Improved
Add GCT-B1 channel attentionResNet101 · ImageNet validationTop-1 accuracy: 77.37 -> 78.85Improved
Add GCT-B0 channel attentionShufflenetV2 · ImageNet validationTop-1 accuracy: 69.36 -> 70.79Improved
Add GCT-B0 channel attentionFaster R-CNN ResNet50 · COCO 2017 validationAPbbox 0.5:0.95: 37.4 -> 38.8Improved
Add GCT-B1 channel attentionFaster R-CNN ResNet50 · COCO 2017 validationAPbbox 0.5:0.95: 37.4 -> 38.9Improved
Add GCT-B0 channel attentionMask R-CNN ResNet50 · COCO 2017 validationAPbbox 0.5:0.95: 38.2 -> 39.3Improved
Add GCT-B0 channel attentionRetinaNet ResNet50 · COCO 2017 validationAPbbox 0.5:0.95: 36.5 -> 38.1Improved
Add GCT-B0 channel attentionMask R-CNN ResNet50 · COCO 2017 validationAPmask 0.5:0.95: 34.7 -> 35.8Improved
Vary fixed Gaussian standard deviation c from 1 to 6; c=2 gives the best reported score (ablation)GCT-ResNet50-B0 · ImageNet validationTop-1 accuracy: 77.15, 77.51, 77.24, 77.27, 77.11, 77.06Mixed

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