Swin-Tiny: Optimized for Qualcomm Devices

SwinTiny 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-Tiny 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-Tiny 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-Tiny 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): 110 MB
• Model size (w8a16): 29.9 MB
• Number of parameters: 28.8M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Tiny	ONNX	float	Snapdragon® X2 Elite	3.936 ms	2 - 2 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® X Elite	9.486 ms	58 - 58 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.925 ms	0 - 264 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Gen 1 Mobile	13.826 ms	1 - 257 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	8.913 ms	0 - 4 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	8.922 ms	0 - 65 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS8450	13.826 ms	1 - 257 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	13.989 ms	1 - 4 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	9.486 ms	58 - 58 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® Dragonwing™ Q-8750	4.502 ms	0 - 442 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Elite Mobile	4.502 ms	0 - 442 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.692 ms	0 - 448 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® X2 Elite	3.469 ms	1 - 1 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® X Elite	8.226 ms	31 - 31 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	5.166 ms	0 - 332 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	9.717 ms	0 - 278 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	25.198 ms	0 - 3 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	7.08 ms	0 - 4 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	7.838 ms	0 - 35 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS8450	9.717 ms	0 - 278 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	8.14 ms	0 - 3 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	8.226 ms	31 - 31 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	35.581 ms	2 - 394 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	8.25 ms	0 - 235 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	3.981 ms	0 - 265 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Elite Mobile	3.981 ms	0 - 265 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	3.266 ms	0 - 273 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	8.25 ms	0 - 235 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® X2 Elite	4.267 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® X Elite	9.856 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.96 ms	0 - 246 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	14.5 ms	0 - 240 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	9.025 ms	1 - 4 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	21.244 ms	1 - 175 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	9.105 ms	1 - 172 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA8775P	10.41 ms	1 - 435 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA8650P	10.41 ms	1 - 435 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA8255P	10.41 ms	1 - 435 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS8450	14.5 ms	0 - 240 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	10.262 ms	3 - 5 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	9.856 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	4.514 ms	0 - 415 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA7255P	21.244 ms	1 - 175 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA8295P	13.831 ms	0 - 433 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.514 ms	0 - 415 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.775 ms	1 - 415 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® X2 Elite	4.412 ms	0 - 0 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® X Elite	10.98 ms	0 - 0 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	6.597 ms	0 - 508 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	8.838 ms	0 - 3 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	17.784 ms	0 - 437 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.201 ms	0 - 3 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® SA8775P	10.654 ms	0 - 443 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® SA8650P	10.654 ms	0 - 443 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® SA8255P	10.654 ms	0 - 443 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	10.513 ms	0 - 2 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	10.98 ms	0 - 0 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	40.504 ms	0 - 430 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	10.643 ms	0 - 434 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	4.892 ms	0 - 430 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® SA7255P	17.784 ms	0 - 437 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	4.892 ms	0 - 430 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	3.891 ms	0 - 437 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	10.643 ms	0 - 434 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	6.038 ms	0 - 263 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	14.287 ms	0 - 257 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	9.062 ms	0 - 61 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	21.19 ms	0 - 200 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	9.243 ms	0 - 3 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8775P	10.413 ms	0 - 201 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8650P	10.413 ms	0 - 201 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8255P	10.413 ms	0 - 201 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS8450	14.287 ms	0 - 257 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	10.182 ms	0 - 60 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	4.649 ms	0 - 206 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA7255P	21.19 ms	0 - 200 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8295P	13.855 ms	0 - 199 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Elite Mobile	4.649 ms	0 - 206 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.833 ms	0 - 214 MB	NPU

License

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