NASNet: Optimized for Qualcomm Devices

NASNet is a CNN-based architecture discovered via Neural Architecture Search (NAS) that can classify images from the Imagenet dataset.

This is based on the implementation of NASNet 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	w8a8_mixed_fp16	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit NASNet 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 NASNet on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: nasnetalarge.tf_in1k
• Input resolution: 224x224
• GMACs: 5.9
• Activations (M): 19.4
• Number of parameters: 88.7M
• Model size (float): 338 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
NASNet	ONNX	float	Snapdragon® X2 Elite	9.628 ms	179 - 179 MB	NPU
NASNet	ONNX	float	Snapdragon® X Elite	18.994 ms	188 - 188 MB	NPU
NASNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	14.224 ms	1 - 560 MB	NPU
NASNet	ONNX	float	Snapdragon® 8 Gen 1 Mobile	52.539 ms	2 - 509 MB	NPU
NASNet	ONNX	float	Qualcomm® QCS8550 (Proxy)	18.865 ms	0 - 194 MB	NPU
NASNet	ONNX	float	Qualcomm® QCS8450	52.539 ms	2 - 509 MB	NPU
NASNet	ONNX	float	Snapdragon® 8 Elite Mobile	11.763 ms	1 - 403 MB	NPU
NASNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	9.431 ms	1 - 411 MB	NPU
NASNet	ONNX	float	Qualcomm® QCS9075	29.458 ms	1 - 46 MB	NPU
NASNet	ONNX	float	Qualcomm® QCS8750	11.763 ms	1 - 403 MB	NPU
NASNet	ONNX	float	Qualcomm® QCS7181	18.994 ms	188 - 188 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® X2 Elite	4.622 ms	180 - 180 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® X Elite	9.438 ms	149 - 149 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	6.51 ms	0 - 609 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® 8 Gen 1 Mobile	19.186 ms	1 - 593 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Qualcomm® QCS8550 (Proxy)	9.226 ms	1 - 162 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Qualcomm® QCS8450	19.186 ms	1 - 593 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Qualcomm® QCS9075	10.017 ms	0 - 46 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® 8 Elite Mobile	5.311 ms	1 - 487 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Snapdragon® 8 Elite Gen 5 Mobile	4.713 ms	0 - 478 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Qualcomm® QCS8750	5.311 ms	1 - 487 MB	NPU
NASNet	ONNX	w8a8_mixed_fp16	Qualcomm® QCS7181	9.438 ms	149 - 149 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® X2 Elite	8.842 ms	1 - 1 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® X Elite	16.703 ms	1 - 1 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	12.088 ms	0 - 515 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	48.588 ms	0 - 472 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS8275	89.226 ms	1 - 351 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	16.242 ms	1 - 4 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS8450	48.588 ms	0 - 472 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	9.757 ms	1 - 351 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® SA8295P	38.656 ms	1 - 310 MB	NPU
NASNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	7.732 ms	0 - 367 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® SA7255P	89.226 ms	1 - 351 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS9075	26.592 ms	3 - 6 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS8750	9.757 ms	1 - 351 MB	NPU
NASNet	QNN_DLC	float	Qualcomm® QCS7181	16.703 ms	1 - 1 MB	NPU
NASNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	12.677 ms	0 - 695 MB	NPU
NASNet	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	50.114 ms	0 - 652 MB	NPU
NASNet	TFLITE	float	Qualcomm® QCS8275	89.665 ms	0 - 538 MB	NPU
NASNet	TFLITE	float	Qualcomm® QCS8550 (Proxy)	16.786 ms	0 - 197 MB	NPU
NASNet	TFLITE	float	Qualcomm® SA8775P	128.657 ms	0 - 33 MB	GPU
NASNet	TFLITE	float	Qualcomm® SA8650P	128.657 ms	0 - 33 MB	GPU
NASNet	TFLITE	float	Qualcomm® SA8255P	128.657 ms	0 - 33 MB	GPU
NASNet	TFLITE	float	Qualcomm® QCS8450	50.114 ms	0 - 652 MB	NPU
NASNet	TFLITE	float	Snapdragon® 8 Elite Mobile	10.319 ms	0 - 533 MB	NPU
NASNet	TFLITE	float	Qualcomm® SA8295P	40.092 ms	0 - 486 MB	NPU
NASNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	8.164 ms	0 - 551 MB	NPU
NASNet	TFLITE	float	Qualcomm® SA7255P	89.665 ms	0 - 538 MB	NPU
NASNet	TFLITE	float	Qualcomm® QCS9075	25.195 ms	0 - 192 MB	NPU
NASNet	TFLITE	float	Qualcomm® QCS8750	10.319 ms	0 - 533 MB	NPU

License

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

References

• Learning Transferable Architectures for Scalable Image Recognition
• 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.