AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

FastViT: A Fast Hybrid Vision Transformer Using Structural Reparameterization

Pavan Kumar Anasosalu Vasu, James Gabriel, Jeff Zhu, Oncel Tuzel, Anurag Ranjan

ICCV 2023 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 278 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
replace pooling with RepMixer (ablation)PoolFormer-S12 · ImageNet-1kTop-1 accuracy / mobile latency: 77.6 -> 78.5; latency 2.25 -> 1.58 msImproved
factorize dense kxk convolutions (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 78.5 -> 78.0Hurt
add linear train-time overparameterization (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 78.0 -> 78.9Improved
add train-time overparameterization (ablation)FastViT-SA12 · ImageNet-1kTop-1 accuracy: 80.0 -> 80.6Improved
add train-time overparameterization (ablation)FastViT-SA36 · ImageNet-1kTop-1 accuracy: 83.3 -> 83.6Improved
also overparameterize ConvFFN layers (ablation)FastViT-SA12 · ImageNet-1kTop-1 accuracy / train time: 80.6 -> 80.6; train time 33.4 -> 40.1 hrsNo clear effect
add large kernel conv in FFN (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 78.9 -> 79.4Improved
add large kernel conv in patch embedding (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 79.4 -> 79.8Improved
large kernel convs instead of self-attention in last stage (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy / latency: V5 80.6 at 1.6 ms vs V3 81.0 at 3.7 msMixed
increase kernel size beyond 7x7 (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 79.8 -> 79.6 (9x9), 79.8 (11x11)No clear effect
replace BatchNorm with LayerNorm (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy / latency: 79.8 -> 79.7; latency 1.4 -> 1.7 msHurt
replace GELU with ReLU (ablation)FastViT-S12 · ImageNet-1kTop-1 accuracy: 79.8 -> 79.4Hurt

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