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 2.45, 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-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® X2 Elite	8.411 ms	2 - 2 MB	NPU
Swin-Base	ONNX	float	Snapdragon® X Elite	19.492 ms	176 - 176 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Gen 3 Mobile	12.791 ms	0 - 535 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Gen 1 Mobile	29.332 ms	1 - 525 MB	NPU
Swin-Base	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	18.815 ms	0 - 195 MB	NPU
Swin-Base	ONNX	float	Qualcomm® QCS8450	29.332 ms	1 - 525 MB	NPU
Swin-Base	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	22.055 ms	0 - 4 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	7.958 ms	1 - 411 MB	NPU
Swin-Base	ONNX	float	Snapdragon® 8 Elite Mobile	9.656 ms	1 - 387 MB	NPU
Swin-Base	ONNX	float	Qualcomm® Dragonwing™ Q-8750	9.656 ms	1 - 387 MB	NPU
Swin-Base	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	19.492 ms	176 - 176 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® X2 Elite	6.964 ms	1 - 1 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® X Elite	16.304 ms	91 - 91 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	10.465 ms	0 - 540 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.345 ms	0 - 5 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	16.527 ms	0 - 3 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	17.802 ms	0 - 621 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	6.528 ms	0 - 433 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	114.158 ms	0 - 691 MB	NPU
Swin-Base	ONNX	w8a16	Snapdragon® 8 Elite Mobile	8.188 ms	0 - 421 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	17.802 ms	0 - 621 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	8.188 ms	0 - 421 MB	NPU
Swin-Base	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	16.304 ms	91 - 91 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® X2 Elite	8.467 ms	1 - 1 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® X Elite	19.28 ms	1 - 1 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	12.432 ms	0 - 515 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	29.076 ms	0 - 501 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS8275	53.624 ms	1 - 363 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	18.441 ms	1 - 3 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8775P	21.361 ms	1 - 362 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8650P	21.361 ms	1 - 362 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8255P	21.361 ms	1 - 362 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® QCS8450	29.076 ms	0 - 501 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	21.8 ms	3 - 5 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	7.555 ms	0 - 385 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA7255P	53.624 ms	1 - 363 MB	NPU
Swin-Base	QNN_DLC	float	Snapdragon® 8 Elite Mobile	9.394 ms	1 - 366 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® SA8295P	27.651 ms	1 - 352 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	9.394 ms	1 - 366 MB	NPU
Swin-Base	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	19.28 ms	1 - 1 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® X2 Elite	8.43 ms	0 - 0 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® X Elite	20.526 ms	0 - 0 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	12.777 ms	0 - 507 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® QCS8275	35.354 ms	0 - 412 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	19.188 ms	0 - 3 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA8775P	19.651 ms	0 - 412 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA8650P	19.651 ms	0 - 412 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA8255P	19.651 ms	0 - 412 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	20.129 ms	2 - 4 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	21.543 ms	2 - 607 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	7.627 ms	0 - 421 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	120.505 ms	1 - 935 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® SA7255P	35.354 ms	0 - 412 MB	NPU
Swin-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	9.566 ms	0 - 406 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	21.543 ms	2 - 607 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	9.566 ms	0 - 406 MB	NPU
Swin-Base	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	20.526 ms	0 - 0 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	12.742 ms	1 - 1052 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	28.953 ms	0 - 510 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS8275	53.501 ms	0 - 710 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	18.991 ms	0 - 4 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8775P	21.955 ms	0 - 374 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8650P	21.955 ms	0 - 374 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8255P	21.955 ms	0 - 374 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® QCS8450	28.953 ms	0 - 510 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	22.38 ms	0 - 178 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	7.939 ms	0 - 400 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA7255P	53.501 ms	0 - 710 MB	NPU
Swin-Base	TFLITE	float	Snapdragon® 8 Elite Mobile	9.735 ms	0 - 387 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® SA8295P	28.307 ms	0 - 371 MB	NPU
Swin-Base	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	9.735 ms	0 - 387 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.