AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Blending Anti-Aliasing into Vision Transformer

Shengju Qian, Hao Shao, Yi Zhu, Mu Li, Jiaya Jia

NeurIPS 2021 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 0 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 filter-bank anti-aliasing (ARM) to attention maps in early layersSwin-T · ImageNet-1ktop-1 accuracy: 81.2 -> 82.0Improved
Add ARM to other transformer families (DeiT-S, DeiT-B, CoaT, T2T)DeiT/CoaT/T2T · ImageNet-1ktop-1 accuracy: 79.8 -> 80.7Improved
Use fixed Gaussian filter (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.2 -> 81.5Improved
Use learnable convolution filter instead of Gaussian (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.5 -> 81.6No clear effect
Use pre-defined filter bank (n=8) instead of Gaussian (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.5 -> 82.0Improved
Apply anti-aliasing after patch embedding (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.2 -> 81.0Hurt
Apply anti-aliasing to attention maps instead of patch embedding (ablation)DeiT-S · ImageNet-1ktop-1 accuracy: 79.8 -> 80.7Improved
Apply anti-aliasing after attention plus shortcut (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.2 -> 79.8Hurt
Apply anti-aliasing at patch merging (downsampling) (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.2 -> 79.5Hurt
Filter only layers 1-2 instead of all four layers (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.7 vs 81.3Mixed
Increase filter bank size from 2 to 8 (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 81.6 -> 82.0Improved
Increase filter bank size beyond 8 (16, 24) (ablation)Swin-T · ImageNet-1ktop-1 accuracy: 82.0 -> 81.8No 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