SwinV2-Base: Optimized for Qualcomm Devices

SwinV2Base 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 SwinV2-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
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 SwinV2-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 SwinV2-Base on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: Imagenet
• Input resolution: 256x256
• 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
SwinV2-Base	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	11.438 ms	1 - 422 MB	NPU
SwinV2-Base	QNN_DLC	float	Snapdragon® X2 Elite	12.358 ms	1 - 1 MB	NPU
SwinV2-Base	QNN_DLC	float	Snapdragon® X Elite	29.04 ms	1 - 1 MB	NPU
SwinV2-Base	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	19.599 ms	0 - 546 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS8275	74.007 ms	1 - 394 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	27.979 ms	1 - 2 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® SA8775P	31.572 ms	1 - 388 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® SA8650P	31.572 ms	1 - 388 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® SA8255P	31.572 ms	1 - 388 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	41.545 ms	0 - 534 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® SA7255P	74.007 ms	1 - 394 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® SA8295P	37.877 ms	1 - 378 MB	NPU
SwinV2-Base	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	14.616 ms	1 - 390 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS9075	36.634 ms	1 - 3 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS8750	14.616 ms	1 - 390 MB	NPU
SwinV2-Base	QNN_DLC	float	Qualcomm® QCS7181	29.04 ms	1 - 1 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	11.379 ms	0 - 942 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Snapdragon® X2 Elite	12.161 ms	0 - 0 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Snapdragon® X Elite	30.804 ms	0 - 0 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	19.666 ms	0 - 2030 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® QCS8275	52.495 ms	0 - 910 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	29.149 ms	0 - 3 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® SA8775P	29.717 ms	0 - 907 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® SA8650P	29.717 ms	0 - 907 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® SA8255P	29.717 ms	0 - 907 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® SA7255P	52.495 ms	0 - 910 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	14.721 ms	0 - 912 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® QCS9075	34.328 ms	0 - 2 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® QCS8750	14.721 ms	0 - 912 MB	NPU
SwinV2-Base	QNN_DLC	w8a16	Qualcomm® QCS7181	30.804 ms	0 - 0 MB	NPU
SwinV2-Base	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	11.388 ms	0 - 932 MB	NPU
SwinV2-Base	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	20.151 ms	0 - 2185 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® QCS8275	71.155 ms	0 - 885 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® QCS8550 (Proxy)	29.41 ms	0 - 5 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® SA8775P	32.509 ms	0 - 878 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® SA8650P	32.509 ms	0 - 878 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® SA8255P	32.509 ms	0 - 878 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® QCS8450 (Proxy)	43.111 ms	0 - 679 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® SA7255P	71.155 ms	0 - 885 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® SA8295P	40.548 ms	0 - 875 MB	NPU
SwinV2-Base	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	14.935 ms	0 - 895 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® QCS9075	36.338 ms	0 - 180 MB	NPU
SwinV2-Base	TFLITE	float	Qualcomm® QCS8750	14.935 ms	0 - 895 MB	NPU

License

• The license for the original implementation of SwinV2-Base can be found here.

References

• Swin Transformer V2: Scaling Up Capacity and Resolution
• 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.