Swin-Base: Optimized for Qualcomm Devices

SwinBase is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This is based on the implementation of Swin-Base found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime	Precision	Chipset	SDK Versions	Download
ONNX	float	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Swin-Base on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

• Custom weights (e.g., fine-tuned checkpoints)
• Custom input shapes
• Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for Swin-Base on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: Imagenet
• Input resolution: 224x224
• Number of parameters: 88.8M
• Model size (float): 339 MB
• Model size (w8a16): 90.2 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Base	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	8.962 ms	1 - 539 MB	NPU
Swin-Base	ONNX	float	Snapdragon® X2 Elite	9.862 ms	187 - 187 MB	NPU
Swin-Base	ONNX	float	Snapdragon® X Elite	24.088 ms	186 - 186 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Gen 3 Mobile	15.991 ms	0 - 594 MB	NPU
Swin-Base	ONNX	float	Qualcomm® QCS8550 (Proxy)	22.958 ms	0 - 197 MB	NPU
Swin-Base	ONNX	float	Qualcomm® QCS9075	27.442 ms	0 - 4 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	11.706 ms	1 - 520 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	7.166 ms	0 - 475 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® X2 Elite	7.739 ms	99 - 99 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® X Elite	18.014 ms	106 - 106 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	11.731 ms	0 - 598 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® QCS6490	1121.83 ms	53 - 83 MB	CPU
Swin-Base	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	17.208 ms	0 - 128 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® QCS9075	21.312 ms	0 - 3 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® QCM6690	632.98 ms	121 - 139 MB	CPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	9.178 ms	0 - 470 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	598.327 ms	145 - 168 MB	CPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	7.949 ms	0 - 398 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® X2 Elite	8.881 ms	1 - 1 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® X Elite	20.511 ms	1 - 1 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	13.446 ms	0 - 527 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	55.49 ms	1 - 373 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	19.69 ms	1 - 3 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8775P	22.553 ms	1 - 371 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS9075	24.09 ms	1 - 3 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	30.318 ms	0 - 519 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA7255P	55.49 ms	1 - 373 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8295P	28.867 ms	0 - 362 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	9.739 ms	1 - 369 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	7.609 ms	0 - 443 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® X2 Elite	8.571 ms	0 - 0 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® X Elite	21.029 ms	0 - 0 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	13.125 ms	0 - 536 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® QCS8275 (Proxy)	36.174 ms	0 - 435 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	19.627 ms	0 - 404 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA8775P	20.007 ms	0 - 436 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® QCS9075	23.56 ms	0 - 2 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® QCM6690	124.594 ms	0 - 901 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA7255P	36.174 ms	0 - 435 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	9.805 ms	0 - 426 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	22.046 ms	0 - 823 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	8.724 ms	0 - 409 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	14.988 ms	0 - 535 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS8275 (Proxy)	58.496 ms	0 - 387 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS8550 (Proxy)	21.905 ms	0 - 4 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8775P	25.026 ms	0 - 386 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS9075	27.434 ms	0 - 178 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS8450 (Proxy)	32.232 ms	0 - 531 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA7255P	58.496 ms	0 - 387 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8295P	31.877 ms	0 - 388 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	11.147 ms	0 - 364 MB	NPU

License

• The license for the original implementation of Swin-Base can be found here.

References

• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Source Model Implementation

Community

• Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
• For questions or feedback please reach out to us.