AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

A Survey on Efficient Vision Transformers: Algorithms, Techniques, and Performance Benchmarking

Lorenzo Papa, Paolo Russo, Irene Amerini, Luping Zhou

TPAMI 2024 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 213 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
apply Castling linear attention trainingMViTv2-T · ImageNet-1k classification (CA models)Top-1 (%): 82.3 -> 84.1Improved
apply Castling linear attention trainingMViTv2-S · ImageNet-1k classification (CA models)Top-1 (%): 83.6 -> 84.6Improved
apply Castling linear attention trainingMViTv2-B · ImageNet-1k classification (CA models)Top-1 (%): 84.4 -> 85.0Improved
scale from Ortho-T to Ortho-SOrtho · ImageNet-1k classification (CA models)Top-1 (%): 74.0 -> 83.4Improved
scale from Ortho-S to Ortho-LOrtho · ImageNet-1k classification (CA models)Top-1 (%): 83.4 -> 84.2Improved
use UperNet instead of Semantic FPN decoderOrtho-S · ADE20K semantic segmentation (CA encoders)mIoU (%): 48.2 -> 48.5Improved
use UperNet instead of Semantic FPN decoderOrtho-B · ADE20K semantic segmentation (CA encoders)mIoU (%): 49.0 -> 49.8Improved
replace SimA with 8-token patching (T12/16 to T12/8)XCiT SimA · ImageNet-1k classification (CA models)Top-1 (%): 82.1 -> 79.4Hurt
scale from MViTv2-T to MViTv2-LMViTv2 · COCO val2017 detection (CA encoders)AP box: 48.2 -> 51.8Improved
scale from PVT-Tiny to PVT-LargePVT · COCO val2017 detection (CA encoders)AP box: 36.7 -> 42.9Improved

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