Research map / Architectures and efficiency / Attention and transformers
SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applications
Abdelrahman Shaker, Muhammad Maaz, Hanoona Rasheed, Salman Khan, Ming-Hsuan Yang, Fahad Shahbaz Khan
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 pool mixers with the proposed Conv. Encoder (ablation) | EfficientFormer-L1 · ImageNet-1K val | top-1 accuracy (%): 79.2 -> 79.9 | Improved |
| replace self-attention with efficient additive attention (ablation) | EfficientFormer-L1 · ImageNet-1K val | top-1 accuracy (%): 79.9 -> 79.7 top-1; latency 1.1 -> 1.0 ms | Mixed |
| use SwiftFormer encoder (additive attention) in all stages (ablation) | SwiftFormer-L1 · ImageNet-1K val | top-1 accuracy (%): 79.7 -> 80.9 | Improved |
| SwiftFormer-XS versus MobileViT-XS and EdgeViT-XXS | SwiftFormer-XS · ImageNet-1K val | latency iPhone 14 (ms): 0.7 vs 1.5 ms (MobileViT-XS); top-1 75.7 vs 74.8 | Improved |
| SwiftFormer-L3 versus EfficientFormer-L3 | SwiftFormer-L3 · ImageNet-1K val | top-1 accuracy (%): 82.4 -> 83.0 | Improved |
| SwiftFormer backbone versus EfficientFormer-L1 on COCO and ADE20K | SwiftFormer-L1 · COCO val / ADE20K val | AP box / mIoU: 37.9 -> 41.2 AP box; 38.9 -> 41.4 mIoU | Improved |
| SwiftFormer-L1 versus fully convolutional MobileNet-v3 on A100 throughput | SwiftFormer-L1 · ImageNet-1K val | throughput A100 (img/s): 4469 vs 10351 | Hurt |
Closest papers by meaning
- Mobile Attention: Mobile-Friendly Linear-Attention for Vision Transformers (ICML 2024 · similarity 0.90)
- EfficientFormer: Vision Transformers at MobileNet Speed (NeurIPS 2022 · similarity 0.88)
- PADRe: A Unifying Polynomial Attention Drop-in Replacement for Efficient Vision Transformer (ICLR 2025 · similarity 0.87)
- SHViT: Single-Head Vision Transformer with Memory Efficient Macro Design (CVPR 2024 · similarity 0.86)
- IFORMER: INTEGRATING CONVNET AND TRANSFORMER FOR MOBILE APPLICATION (ICLR 2025 · similarity 0.86)
- Real-time Core-Periphery Guided ViT with Smart Data Layout Selection on Mobile Devices (NeurIPS 2024 · similarity 0.86)
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.
