AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

MetaFormer Baselines for Vision

Weihao Yu, Chenyang Si, Pan Zhou, Mi Luo, Yichen Zhou, Jiashi Feng, Shuicheng Yan, Xinchao Wang

TPAMI 2023 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 418 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
use random mixing in the top two stages instead of identity mappingIdentityFormer-S12 · ImageNet-1K valtop-1 accuracy: 74.6 -> 76.6Improved
scale identity-mapping MetaFormer up to M48IdentityFormer · ImageNet-1K valtop-1 accuracy: 74.6 -> 80.4Improved
use pooling instead of random mixingRandFormer-M48 · ImageNet-1K valtop-1 accuracy: 81.4 -> 82.6Improved
add a 4-layer conv stem (ablation)isotropic IdentityFormer-S · ImageNet-1K valtop-1 accuracy: 68.2 -> 75.4Improved
replace conv token mixer with attention in top two stages (CAFormer vs ConvFormer)ConvFormer-S18 · ImageNet-1K valtop-1 accuracy: 83.0 -> 83.6Improved
add ImageNet-21K pretrainingCAFormer-B36 · ImageNet-1K valtop-1 accuracy: 85.5 -> 87.4Improved
use StarReLU instead of ReLU (ablation)ConvFormer-S18 · ImageNet-1K valtop-1 accuracy: 82.1 -> 83.0Improved
use StarReLU instead of GELU (ablation)ConvFormer-S18 · ImageNet-1K valtop-1 accuracy: 82.7 -> 83.0Improved
use StarReLU instead of GELU (ablation)CAFormer-S18 · ImageNet-1K valtop-1 accuracy: 83.4 -> 83.6Improved
use StarReLU instead of Squared ReLU (ablation)ConvFormer-S18 · ImageNet-1K valtop-1 accuracy: 82.6 -> 83.0Improved
use ResScale instead of LayerScale (ablation)CAFormer-S18 · ImageNet-1K valtop-1 accuracy: 83.0 -> 83.6Improved
disable biases in blocks (ablation)ConvFormer-S18 · ImageNet-1K valtop-1 accuracy: 83.0 -> 83.0No 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