EasyDeepLearn

What does Swin Transformer do differently from ViT?

hard

Answer

  • Swin uses hierarchical (multi-scale) feature maps like CNNs, with attention computed inside local windows (e.g., 7×7 patches).
  • Adjacent transformer blocks shift the windows so information flows across window boundaries — hence 'Swin' (Shifted Windows).
  • O(N) complexity in image size vs ViT's O(N2)O(N^{2}), and produces multi-scale features suitable for detection / segmentation.
Check yourself — multiple choice
  • Global attention like ViT
  • Local windowed attention + shifted windows across layers → hierarchical, O(N) in image size
  • No positional info
  • Only for text

Swin: shifted windows + hierarchical maps → linear complexity, dense-prediction friendly.

#transformers#computer-vision#architectures

Practise Deep Learning

214 interview questions in this topic.

Related questions