Swin-Small: Optimized for Qualcomm Devices

SwinSmall 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-Small 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 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	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-Small 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-Small 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: 50.4M
• Model size (float): 193 MB
• Model size (w8a16): 52.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Small	ONNX	float	Snapdragon® X2 Elite	6.781 ms	212 - 212 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X Elite	16.243 ms	148 - 148 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.25 ms	1 - 411 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 1 Mobile	23.937 ms	1 - 402 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8550 (Proxy)	15.556 ms	0 - 115 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8450	23.937 ms	1 - 402 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite Mobile	7.835 ms	1 - 606 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.386 ms	1 - 616 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS9075	19.721 ms	1 - 46 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8750	7.835 ms	1 - 606 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS7181	16.243 ms	148 - 148 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X2 Elite	5.986 ms	212 - 212 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X Elite	13.857 ms	149 - 149 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	8.708 ms	0 - 520 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	13.154 ms	0 - 116 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS9075	15.981 ms	0 - 45 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	5.637 ms	0 - 449 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Mobile	6.918 ms	0 - 441 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	13.947 ms	0 - 387 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCM6690	65.285 ms	0 - 555 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS7790	13.947 ms	0 - 387 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8750	6.918 ms	0 - 441 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS7181	13.857 ms	149 - 149 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X2 Elite	7.149 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X Elite	16.31 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.904 ms	0 - 406 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	23.976 ms	0 - 398 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8275	37.947 ms	1 - 275 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	15.357 ms	1 - 308 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8450	23.976 ms	0 - 398 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Mobile	7.596 ms	1 - 588 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8295P	22.669 ms	1 - 270 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	6.227 ms	0 - 600 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA7255P	37.947 ms	1 - 275 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS9075	19.524 ms	3 - 5 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8750	7.596 ms	1 - 588 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS7181	16.31 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X2 Elite	7.623 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X Elite	17.398 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	10.731 ms	0 - 868 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8275	29.302 ms	0 - 636 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	16.21 ms	0 - 268 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS9075	18.882 ms	2 - 4 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	6.458 ms	0 - 640 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	8.08 ms	0 - 622 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	17.063 ms	0 - 652 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCM6690	69.138 ms	0 - 751 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA7255P	29.302 ms	0 - 636 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS7790	17.063 ms	0 - 652 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8750	8.08 ms	0 - 622 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS7181	17.398 ms	0 - 0 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.144 ms	0 - 429 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	23.361 ms	0 - 410 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8275	37.401 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8550 (Proxy)	15.448 ms	0 - 3 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8775P	70.422 ms	0 - 33 MB	GPU
Swin-Small	TFLITE	float	Qualcomm® SA8650P	70.422 ms	0 - 33 MB	GPU
Swin-Small	TFLITE	float	Qualcomm® SA8255P	70.422 ms	0 - 33 MB	GPU
Swin-Small	TFLITE	float	Qualcomm® QCS8450	23.361 ms	0 - 410 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Mobile	7.84 ms	0 - 304 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8295P	22.978 ms	0 - 297 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	6.579 ms	0 - 322 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA7255P	37.401 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS9075	20.015 ms	0 - 104 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8750	7.84 ms	0 - 304 MB	NPU

License

• The license for the original implementation of Swin-Small 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.