AI Paper Atlas

Research map / Detection and segmentation / CNN features

VSA: Learning Varied-Size Window Attention in Vision Transformers

Qiming Zhang, Yufei Xu, Jing Zhang, Dacheng Tao

ECCV 2022 · Detection and segmentation › CNN features · approach: Convolution-transformer backbones · 77 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 VSA in place of fixed window attentionSwin-T · ImageNet-1k val (224)top-1 accuracy: 81.2 -> 82.3Improved
add VSA to the model at 384 inputSwin-T · ImageNet-1k val (384)top-1 accuracy: 81.4 -> 83.2Improved
add VSA to the model at 480 inputSwin-T · ImageNet-1k val (480)top-1 accuracy: 81.5 -> 83.4Improved
add VSA to ViTAEv2-SViTAEv2-S · ImageNet-1k val (224)top-1 accuracy: 82.2 -> 82.7Improved
add VSA to Mask RCNN 1x backboneSwin-T / ViTAEv2-S · COCO val Mask RCNN 1xbox AP: 43.7 -> 45.6Improved
add VSA to Mask RCNN 3x backboneSwin-T / ViTAEv2-S · COCO val Mask RCNN 3xbox AP: 46.0 -> 47.5Improved
add VSA to Cascade RCNN backboneSwin-T · COCO val Cascade RCNNbox AP: 50.2 -> 51.4Improved
add VSA to UperNet backboneSwin-T · Cityscapes 512x1024 40kmIoU: 78.9 -> 80.8Improved
add VSA at stage 4 only (ablation)Swin-T · ImageNet-1k val (ablation)top-1 accuracy: 81.2 -> 81.4Improved
extend VSA from stage 4 to stages 3-4 (ablation)Swin-T · ImageNet-1k val (ablation)top-1 accuracy: 81.4 -> 81.9Improved
extend VSA to all four stages (ablation)Swin-T · ImageNet-1k val (ablation)top-1 accuracy: 82.1 -> 82.3Improved
add VSR (varied-size window regression) alone (ablation)Swin-T · ImageNet-1k val (ablation)top-1 accuracy: 81.2 -> 81.6Improved

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