RegNet: Optimized for Qualcomm Devices

RegNet 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 RegNet 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	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit RegNet 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 RegNet 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: 15.3M
• Model size (float): 58.3 MB
• Model size (w8a8): 15.4 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
RegNet	ONNX	float	Snapdragon® X2 Elite	0.911 ms	2 - 2 MB	NPU
RegNet	ONNX	float	Snapdragon® X Elite	1.772 ms	39 - 39 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.229 ms	0 - 127 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.731 ms	0 - 116 MB	NPU
RegNet	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.746 ms	0 - 42 MB	NPU
RegNet	ONNX	float	Qualcomm® QCS8450	2.731 ms	0 - 116 MB	NPU
RegNet	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.809 ms	1 - 4 MB	NPU
RegNet	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.772 ms	39 - 39 MB	NPU
RegNet	ONNX	float	Qualcomm® Dragonwing™ Q-8750	0.995 ms	0 - 75 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Elite Mobile	0.995 ms	0 - 75 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.837 ms	0 - 77 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® X2 Elite	0.472 ms	1 - 1 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® X Elite	0.928 ms	20 - 20 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.642 ms	0 - 135 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	1.292 ms	0 - 134 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® Dragonwing™ QCS6490	4.221 ms	0 - 3 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.901 ms	0 - 170 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® QCS8450	1.292 ms	0 - 134 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.159 ms	0 - 3 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	0.928 ms	20 - 20 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	0.549 ms	0 - 90 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.549 ms	0 - 90 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.507 ms	0 - 90 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® X2 Elite	1.204 ms	1 - 1 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® X Elite	2.288 ms	1 - 1 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.414 ms	0 - 127 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	3.501 ms	0 - 116 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	9.87 ms	1 - 71 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.055 ms	1 - 2 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8775P	3.275 ms	0 - 73 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8650P	3.275 ms	0 - 73 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8255P	3.275 ms	0 - 73 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® QCS8450	3.501 ms	0 - 116 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	3.044 ms	1 - 3 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	2.288 ms	1 - 1 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.114 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA7255P	9.87 ms	1 - 71 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8295P	3.472 ms	0 - 61 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.114 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.901 ms	1 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.568 ms	0 - 0 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® X Elite	1.072 ms	0 - 0 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.632 ms	0 - 111 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.307 ms	0 - 111 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS6490	3.962 ms	0 - 2 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8275	2.296 ms	0 - 80 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.896 ms	0 - 94 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8775P	1.291 ms	0 - 82 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8650P	1.291 ms	0 - 82 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8255P	1.291 ms	0 - 82 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS8450	1.307 ms	0 - 111 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.114 ms	0 - 2 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	1.072 ms	0 - 0 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-6690	7.137 ms	0 - 199 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-7790	1.178 ms	0 - 80 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	0.486 ms	0 - 78 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA7255P	2.296 ms	0 - 80 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8295P	1.609 ms	0 - 78 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.486 ms	0 - 78 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.443 ms	0 - 83 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.178 ms	0 - 80 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.392 ms	0 - 152 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	3.481 ms	0 - 137 MB	NPU
RegNet	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	9.873 ms	0 - 89 MB	NPU
RegNet	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.032 ms	0 - 2 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8775P	3.285 ms	0 - 91 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8650P	3.285 ms	0 - 91 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8255P	3.285 ms	0 - 91 MB	NPU
RegNet	TFLITE	float	Qualcomm® QCS8450	3.481 ms	0 - 137 MB	NPU
RegNet	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	3.043 ms	0 - 42 MB	NPU
RegNet	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.107 ms	0 - 93 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA7255P	9.873 ms	0 - 89 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8295P	3.467 ms	0 - 72 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Elite Mobile	1.107 ms	0 - 93 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.899 ms	0 - 96 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.524 ms	0 - 118 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	1.118 ms	0 - 116 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS6490	3.846 ms	0 - 22 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8275	1.982 ms	0 - 75 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.719 ms	0 - 2 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8775P	1.109 ms	0 - 79 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8650P	1.109 ms	0 - 79 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8255P	1.109 ms	0 - 79 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS8450	1.118 ms	0 - 116 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	0.936 ms	0 - 22 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-6690	6.644 ms	0 - 192 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-7790	0.988 ms	0 - 73 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	0.435 ms	0 - 81 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA7255P	1.982 ms	0 - 75 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8295P	1.42 ms	0 - 72 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	0.435 ms	0 - 81 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.393 ms	0 - 76 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	0.988 ms	0 - 73 MB	NPU

License

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

References

• Designing Network Design Spaces
• 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.