swin.md

Swin Transformer

• Paper：Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

• Origin Repo：microsoft/Swin-Transformer

• Code：swin.py

• Evaluate Transforms：

  # backend: pil
  # input_size: 224x224
  transforms = T.Compose([
      T.Resize(248, interpolation='bicubic'),
      T.CenterCrop(224),
      T.ToTensor(),
      T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
  ])
  
  # backend: pil
  # input_size: 384x384
  transforms = T.Compose([
      T.Resize((384, 384), interpolation='bicubic'),
      T.ToTensor(),
      T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
  ])

• Model Details：

  Model	Model Name	Params (M)	FLOPs (G)	Top-1 (%)	Top-5 (%)	Pretrained Model
  Swin-tiny	swin_ti	28	4.5	81.19	95.51	Download
  Swin-small	swin_s	50	8.7	83.18	96.24	Download
  Swin-base	swin_b	88	15.4	83.42	96.45	Download
  Swin-base-384	swin_b_384	88	47.1	84.47	96.95	Download

• Citation：

  @article{liu2021Swin,
  title={Swin Transformer: Hierarchical Vision Transformer using Shifted Windows},
  author={Liu, Ze and Lin, Yutong and Cao, Yue and Hu, Han and Wei, Yixuan and Zhang, Zheng and Lin, Stephen and Guo, Baining},
  journal={arXiv preprint arXiv:2103.14030},
  year={2021}
  }