MNASNet05: Optimized for Qualcomm Devices

MNASNet05 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 MNASNet05 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 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	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 MNASNet05 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 MNASNet05 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: 2.21M
• Model size (float): 8.45 MB
• Model size (w8a16): 2.79 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
MNASNet05	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.217 ms	0 - 31 MB	NPU
MNASNet05	ONNX	float	Snapdragon® X2 Elite	0.228 ms	5 - 5 MB	NPU
MNASNet05	ONNX	float	Snapdragon® X Elite	0.632 ms	5 - 5 MB	NPU
MNASNet05	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.331 ms	0 - 48 MB	NPU
MNASNet05	ONNX	float	Qualcomm® QCS8550 (Proxy)	0.488 ms	0 - 9 MB	NPU
MNASNet05	ONNX	float	Qualcomm® QCS9075	0.765 ms	1 - 3 MB	NPU
MNASNet05	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.264 ms	0 - 28 MB	NPU
MNASNet05	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.217 ms	0 - 32 MB	NPU
MNASNet05	ONNX	w8a16	Snapdragon® X2 Elite	0.235 ms	0 - 0 MB	NPU
MNASNet05	ONNX	w8a16	Snapdragon® X Elite	0.641 ms	2 - 2 MB	NPU
MNASNet05	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.35 ms	0 - 40 MB	NPU
MNASNet05	ONNX	w8a16	Qualcomm® QCS6490	27.633 ms	10 - 13 MB	CPU
MNASNet05	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	0.516 ms	0 - 5 MB	NPU
MNASNet05	ONNX	w8a16	Qualcomm® QCS9075	0.686 ms	0 - 3 MB	NPU
MNASNet05	ONNX	w8a16	Qualcomm® QCM6690	10.346 ms	10 - 18 MB	CPU
MNASNet05	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.26 ms	0 - 27 MB	NPU
MNASNet05	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	7.686 ms	13 - 21 MB	CPU
MNASNet05	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.291 ms	1 - 32 MB	NPU
MNASNet05	QNN_DLC	float	Snapdragon® X2 Elite	0.411 ms	1 - 1 MB	NPU
MNASNet05	QNN_DLC	float	Snapdragon® X Elite	0.945 ms	1 - 1 MB	NPU
MNASNet05	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.513 ms	0 - 45 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	2.334 ms	1 - 27 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	0.79 ms	0 - 12 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® SA8775P	1.106 ms	1 - 31 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® QCS9075	0.976 ms	1 - 3 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	1.564 ms	0 - 48 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® SA7255P	2.334 ms	1 - 27 MB	NPU
MNASNet05	QNN_DLC	float	Qualcomm® SA8295P	1.424 ms	1 - 28 MB	NPU
MNASNet05	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.385 ms	1 - 29 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.302 ms	0 - 29 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.413 ms	0 - 0 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® X Elite	0.921 ms	0 - 0 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	0.523 ms	0 - 37 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCS6490	2.207 ms	2 - 4 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCS8275 (Proxy)	1.658 ms	0 - 26 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	0.772 ms	0 - 2 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® SA8775P	0.95 ms	0 - 27 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCS9075	0.917 ms	0 - 2 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCM6690	3.045 ms	0 - 139 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® QCS8450 (Proxy)	0.957 ms	0 - 39 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® SA7255P	1.658 ms	0 - 26 MB	NPU
MNASNet05	QNN_DLC	w8a16	Qualcomm® SA8295P	1.22 ms	0 - 24 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.353 ms	0 - 30 MB	NPU
MNASNet05	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	0.789 ms	0 - 25 MB	NPU
MNASNet05	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.292 ms	0 - 32 MB	NPU
MNASNet05	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.523 ms	0 - 46 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® QCS8275 (Proxy)	2.331 ms	0 - 28 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® QCS8550 (Proxy)	0.8 ms	0 - 1 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® SA8775P	1.106 ms	0 - 31 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® QCS9075	0.986 ms	0 - 8 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® QCS8450 (Proxy)	1.583 ms	0 - 48 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® SA7255P	2.331 ms	0 - 28 MB	NPU
MNASNet05	TFLITE	float	Qualcomm® SA8295P	1.493 ms	0 - 28 MB	NPU
MNASNet05	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.381 ms	0 - 32 MB	NPU

License

• The license for the original implementation of MNASNet05 can be found here.

References

• MnasNet: Platform-Aware Neural Architecture Search for Mobile
• 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.