AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Rotary Position Embedding for Vision Transformer

Byeongho Heo, Song Park, Dongyoon Han, Sangdoo Yun

ECCV 2024 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 140 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 APE with RoPE-Mixed in ViT-BViT-B · ImageNet-1k val, 512x512 testtop-1 accuracy: 80.5 -> 82.9Improved
use mixed learnable-frequency RoPE instead of axial frequency RoPE (ablation)ViT-B · ImageNet-1k val, multi-resolutiontop-1 accuracy: 82.0 -> 82.9 at 512x512Improved
make axial RoPE frequencies learnable without mixing axes (ablation)ViT-S RoPE-Axial · ImageNet-1k val, ViT-Stop-1 accuracy: -0.2 at 224, +1.3 at 384, +1.7 at 512Mixed
add APE on top of RoPE-Mixed in ViT-B (ablation)ViT-B RoPE-Mixed · ImageNet-1k val, 112 vs 512 testtop-1 accuracy gain over APE: 1.1 -> 2.2 at 112; 2.4 -> 1.9 at 512Mixed
add RPB on top of RoPE-Mixed in Swin-B (ablation)Swin-B RoPE-Mixed · ImageNet-1k val, 384x384gain over RPB: 0.9 -> 0.2Hurt
replace APE with RoPE-Mixed in DINO-ViTDet backboneViT-B · MS-COCO detectionbox AP: 49.4 -> 51.2Improved
replace RPB with RoPE-Mixed in DINO-Swin backboneSwin-T · MS-COCO detectionbox AP: 51.3 -> 51.8Improved
replace APE with RoPE-Mixed + APE in UperNet ViT-BViT-B · ADE20k UperNet single-scalemIoU: 47.7 -> 50.0Improved
add RPB to axial RoPE in Swin-B Mask2Former (ablation)Swin-B RoPE-Axial · ADE20k Swin-Mask2FormermIoU: 52.0 -> 50.0Hurt
use RoPE-Mixed instead of ResFormer-S-224 at small resolutionViT-S · ImageNet-1k val, 128x128top-1 accuracy: 71.4 -> 70.6Hurt
use RoPE-Mixed instead of ResFormer-S-224 at large resolutionViT-S · ImageNet-1k val, 512x512top-1 accuracy: 77.7 -> 79.1Improved

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