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 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	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:

• Input resolution: 224x224
• Model checkpoint: Imagenet
• Model size (float): 193 MB
• Model size (w8a16): 52.5 MB
• Number of parameters: 50.4M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Small	ONNX	float	Snapdragon® X2 Elite	6.899 ms	2 - 2 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X Elite	16.107 ms	101 - 101 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.167 ms	0 - 445 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 1 Mobile	23.437 ms	0 - 430 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	15.606 ms	0 - 4 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.465 ms	0 - 123 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8450	23.437 ms	0 - 430 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	17.49 ms	0 - 4 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	16.107 ms	101 - 101 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ Q-8750	7.804 ms	1 - 610 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite Mobile	7.804 ms	1 - 610 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.44 ms	1 - 621 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X2 Elite	5.982 ms	1 - 1 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X Elite	13.825 ms	54 - 54 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	8.688 ms	0 - 526 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	16.636 ms	0 - 435 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	39.901 ms	0 - 3 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	11.842 ms	0 - 4 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	13.148 ms	0 - 59 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8450	16.636 ms	0 - 435 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	13.835 ms	0 - 3 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	13.825 ms	54 - 54 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	67.201 ms	0 - 555 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	13.972 ms	0 - 388 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	6.991 ms	0 - 437 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Mobile	6.991 ms	0 - 437 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	5.635 ms	0 - 450 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	13.972 ms	0 - 388 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X2 Elite	7.107 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X Elite	16.329 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.909 ms	0 - 405 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	24.017 ms	0 - 395 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	15.522 ms	1 - 4 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	37.938 ms	1 - 275 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.323 ms	1 - 2 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8775P	17.405 ms	1 - 592 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8650P	17.405 ms	1 - 592 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8255P	17.405 ms	1 - 592 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8450	24.017 ms	0 - 395 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	17.542 ms	1 - 3 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	16.329 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	7.674 ms	1 - 587 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA7255P	37.938 ms	1 - 275 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8295P	22.693 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Mobile	7.674 ms	1 - 587 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	6.245 ms	0 - 600 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X2 Elite	7.2 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X Elite	17.395 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	10.746 ms	0 - 868 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	14.376 ms	0 - 3 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	29.247 ms	0 - 635 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	16.178 ms	0 - 2 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8775P	16.77 ms	0 - 638 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8650P	16.77 ms	0 - 638 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8255P	16.77 ms	0 - 638 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	16.649 ms	0 - 2 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	17.395 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	65.485 ms	0 - 751 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	16.976 ms	0 - 652 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	8.228 ms	0 - 622 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA7255P	29.247 ms	0 - 635 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	8.228 ms	0 - 622 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	6.493 ms	0 - 642 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	16.976 ms	0 - 652 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.174 ms	0 - 420 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	23.281 ms	0 - 412 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	15.562 ms	0 - 104 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	37.425 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.478 ms	0 - 3 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8775P	17.448 ms	0 - 302 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8650P	17.448 ms	0 - 302 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8255P	17.448 ms	0 - 302 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8450	23.281 ms	0 - 412 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	17.865 ms	0 - 104 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	7.81 ms	0 - 303 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA7255P	37.425 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8295P	23.0 ms	0 - 297 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Mobile	7.81 ms	0 - 303 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	6.597 ms	0 - 320 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.