Research map / Architectures and efficiency / Attention and transformers
ScalableViT: Rethinking the Context-Oriented Generalization of Vision Transformer
Rui Yang, Hailong Ma, Jie Wu, Yansong Tang, Xuefeng Xiao, Min Zheng, Xiu Li
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 |
|---|---|---|---|
| add SSA instead of WSA in all blocks (ablation) | ScalableViT-S · ADE20K Semantic FPN | mIoU: 38.9 -> 44.4 | Improved |
| add IWSA (LIM) instead of WSA in all blocks (ablation) | ScalableViT-S · ADE20K Semantic FPN | mIoU: 38.9 -> 43.5 | Improved |
| replace WSA with IWSA or SSA (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.4 -> 82.9 | Improved |
| add channel scalability (r_c>1) to SSA (ablation) | ScalableViT-S · ADE20K Semantic FPN | mIoU: 43.7 -> 44.4 | Improved |
| add channel scalability (r_c>1) to SSA (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.6 -> 82.9 | Improved |
| order IWSA before SSA rather than SSA before IWSA (ablation) | ScalableViT-S · ImageNet-1K | top-1: 83.0 -> 83.1 | Improved |
| use IWSA instead of SWSA alternated with SSA (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.9 -> 83.1 | Improved |
| add LIM to baseline (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.7 -> 83.0 | Improved |
| add LEM to baseline (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.7 -> 82.8 | Improved |
| add PEG to baseline (ablation) | ScalableViT-S · ImageNet-1K | top-1: 82.7 -> 82.9 | Improved |
| add LIM to baseline (ablation) | ScalableViT-S · ADE20K Semantic FPN | mIoU: 41.7 -> 43.7 | Improved |
| add PEG together with LIM to baseline (ablation) | ScalableViT-S · ADE20K Semantic FPN | mIoU: 41.7 -> 44.9 | Improved |
Closest papers by meaning
- MaxViT: Multi-axis Vision Transformer (ECCV 2022 · similarity 0.90)
- Efficient Vision Transformers with Partial Attention (ECCV 2024 · similarity 0.90)
- You Only Need Less Attention at Each Stage in Vision Transformers (CVPR 2024 · similarity 0.88)
- Attention Guided CAM: Visual Explanations of Vision Transformer Guided by Self-Attention (AAAI 2024 · similarity 0.87)
- Dissecting Query-Key Interaction in Vision Transformers (NeurIPS 2024 · similarity 0.87)
- VISion On Request: Enhanced VLLM efficiency with sparse, dynamically selected, vision-language interactions (CVPR 2026 · similarity 0.87)
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.
