---
library_name: lucid
license: mit
tags:
  - image-classification
  - swin
  - lucid
datasets:
  - imagenet-1k
pipeline_tag: image-classification
model-index:
  - name: swin-tiny
    results:
      - task: { type: image-classification }
        dataset: { name: ImageNet-1K, type: imagenet-1k }
        metrics:
          - { type: acc@1, value: 81.474 }
          - { type: acc@5, value: 95.776 }
---

# Swin-Tiny

> Liu et al., 2021 — *Swin Transformer: Hierarchical Vision Transformer using Shifted Windows* (arXiv:2103.14030)

[Lucid](https://github.com/ChanLumerico/lucid) port of `torchvision/Swin_T_Weights.IMAGENET1K_V1`,
converted to Lucid-native safetensors.

## Available weights

| Tag | acc@1 | acc@5 | Params | GFLOPs | Size | Source |
|---|---|---|---|---|---|---|
| `IMAGENET1K_V1` *(default)* | 81.474 | 95.776 | 28.3M | 4.491 | 107.93 MB | torchvision |

## Usage

```python
import lucid.models as models
from lucid.models.weights import SwinTinyWeights

# default tag
model = models.swin_tiny_cls(pretrained=True)

# explicit tag (enum or string)
model = models.swin_tiny_cls(weights=SwinTinyWeights.IMAGENET1K_V1)
model = models.swin_tiny_cls(pretrained="IMAGENET1K_V1")

# preprocessing travels with the weights
weights = SwinTinyWeights.IMAGENET1K_V1
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits
```

## Conversion

Converted from `torchvision/Swin_T_Weights.IMAGENET1K_V1` via
`python -m tools.convert_weights swin_tiny --tag IMAGENET1K_V1`.
Key mapping + numerical parity verified against the source.

## License

`mit` — inherited from the original weights.

## Citation

```
@inproceedings{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},
  booktitle={ICCV}, year={2021}
}
```