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 2.50, ONNX Runtime 1.30.0	Download
ONNX	w8a16	Universal	QAIRT 2.50, ONNX Runtime 1.30.0	Download
QNN_DLC	float	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16	Universal	QAIRT 2.50	Download
TFLITE	float	Universal	QAIRT 2.50	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® 8 Elite Gen 5 For Galaxy Mobile	6.403 ms	1 - 301 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.81 ms	1 - 287 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X2 Elite	6.651 ms	2 - 2 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X Elite	15.929 ms	100 - 100 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.146 ms	0 - 419 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 1 Mobile	23.752 ms	1 - 411 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	15.582 ms	0 - 5 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.353 ms	0 - 114 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8450	23.752 ms	1 - 411 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	17.561 ms	0 - 4 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	15.929 ms	100 - 100 MB	NPU
Swin-Small	ONNX	float	Qualcomm® Dragonwing™ Q-8750	7.81 ms	1 - 287 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	5.314 ms	0 - 341 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	6.758 ms	0 - 328 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X2 Elite	5.501 ms	1 - 1 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X Elite	13.946 ms	54 - 54 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	8.952 ms	0 - 443 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	16.634 ms	0 - 443 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	29.267 ms	0 - 4 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	11.808 ms	0 - 5 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	13.306 ms	0 - 59 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8450	16.634 ms	0 - 443 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	14.188 ms	0 - 4 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	13.946 ms	54 - 54 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	58.288 ms	0 - 536 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	14.493 ms	0 - 374 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	6.758 ms	0 - 328 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	14.493 ms	0 - 374 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	6.082 ms	0 - 284 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.489 ms	1 - 272 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X2 Elite	6.779 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X Elite	15.804 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.9 ms	0 - 404 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	23.668 ms	0 - 391 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	15.319 ms	1 - 4 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	14.98 ms	1 - 3 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8775P	16.916 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8650P	16.916 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8255P	16.916 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8450	23.668 ms	0 - 391 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	17.325 ms	1 - 3 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	15.804 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	7.489 ms	1 - 272 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA7255P	37.283 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8295P	22.087 ms	1 - 264 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	6.036 ms	0 - 317 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	7.823 ms	0 - 307 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X2 Elite	6.597 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X Elite	17.088 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	10.807 ms	0 - 409 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	14.289 ms	0 - 4 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	16.13 ms	0 - 242 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8650P	16.763 ms	0 - 638 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8255P	16.763 ms	0 - 638 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	16.875 ms	2 - 5 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	17.088 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	74.064 ms	0 - 525 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	17.129 ms	0 - 351 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	7.823 ms	0 - 307 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA7255P	29.287 ms	0 - 326 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	17.129 ms	0 - 351 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	6.706 ms	0 - 312 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	7.858 ms	0 - 293 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.19 ms	0 - 420 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	23.266 ms	0 - 408 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	15.373 ms	0 - 105 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.438 ms	0 - 4 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8775P	17.395 ms	0 - 293 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8650P	17.395 ms	0 - 293 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8255P	17.395 ms	0 - 293 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8450	23.266 ms	0 - 408 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	17.37 ms	0 - 104 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	7.858 ms	0 - 293 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA7255P	36.861 ms	0 - 292 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8295P	22.673 ms	0 - 286 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.