swin_tiny_patch4_window7_224

Converted TIMM image classification model for LiteRT.

• Source architecture: swin_tiny_patch4_window7_224
• FP32 file: model.tflite
• Static INT8 file: model_static_int8.tflite

Compatibility

File	CPU	GPU	NPU
model.tflite	Supported	Supported with CPU fallback	N/A
model_static_int8.tflite	Supported	Supported with CPU fallback	Qualcomm

The NPU entry denotes compilation support; runtime accuracy is unverified.

model_static_int8.tflite uses INT8 inputs/outputs and convolution/FC quantization; attention, normalization and other operations remain FP32. Apply the source checkpoint’s preprocessing, then quantize the input and dequantize the output using the file’s tensor scales and zero points.

For GPU execution of model_static_int8.tflite, select FP32 GPU precision.

Model Details

• Model Type: Image classification / feature backbone
• Model Stats:
  • Params (M): 28.3
  • GMACs: 4.5
  • Activations (M): 17.1
  • Image size: 224 x 224
• Papers:
  • Swin Transformer: Hierarchical Vision Transformer using Shifted Windows: https://arxiv.org/abs/2103.14030
• Original: https://github.com/microsoft/Swin-Transformer
• Dataset: ImageNet-1k

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={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2021}
}

@misc{rw2019timm,
  author = {Ross Wightman},
  title = {PyTorch Image Models},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  doi = {10.5281/zenodo.4414861},
  howpublished = {\url{https://github.com/huggingface/pytorch-image-models}}
}