Research map / Architectures and efficiency / Attention and transformers
Efficient Vision Transformers with Partial Attention
Xuan-Thuy Vo, Duy-Linh Nguyen, Adri Priadana, Kang-Hyun Jo
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace self-attention with partial attention | DeiT-T · ImageNet-1K | Top-1 accuracy: 72.2 -> 74.2 | Improved |
| replace spatial reduction attention with partial attention | PVT-T · ImageNet-1K | Top-1 accuracy: 75.1 -> 78.5 | Improved |
| replace shifted window attention with partial attention | Swin-T · ImageNet-1K | Top-1 accuracy: 81.3 -> 82.6 | Improved |
| add single-query attention for background tokens (ablation) | PartialFormer-B1 · ImageNet-1K | Top-1 accuracy: 76.5 -> 79.3 | Improved |
| add mixed multi-head self-attention (MMSA) on foreground tokens (ablation) | PartialFormer-B1 · ImageNet-1K | Top-1 accuracy: 77.2 -> 79.3 | Improved |
| add learnable exchange token between foreground and background sets (ablation) | PartialFormer-B1 · ImageNet-1K | Top-1 accuracy: 78.8 -> 79.3 | Improved |
| use adaptive N/NF [64,16,4,1] across stages instead of constant ratio (ablation) | PartialFormer-B1 · ImageNet-1K | Top-1 accuracy: 77.0 -> 78.5 | Improved |
| decrease N/NF from 16 to 4 (more foreground tokens) (ablation) | PartialFormer-B1 · ImageNet-1K | Top-1 accuracy: 77.0 -> 77.3 | Improved |
| use PartialFormer-B3 instead of Swin-T backbone | ADE20K UperNet · ADE20K | mIoU: 44.5 -> 47.0 | Improved |
Closest papers by meaning
- Learned Queries for Efficient Local Attention (CVPR 2022 · similarity 0.91)
- RegionViT: Regional-to-Local Attention for Vision Transformers (ICLR 2022 · similarity 0.90)
- You Only Need Less Attention at Each Stage in Vision Transformers (CVPR 2024 · similarity 0.90)
- Dissecting Query-Key Interaction in Vision Transformers (NeurIPS 2024 · similarity 0.89)
- Skip-Attention: Improving Vision Transformers by Paying Less Attention (ICLR 2024 · similarity 0.89)
- Bridging the Divide: Reconsidering Softmax and Linear Attention (NeurIPS 2024 · similarity 0.88)
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.
