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 2.37, ONNX Runtime 1.23.0	Download
ONNX	w8a8	Universal	QAIRT 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
QNN_DLC	w8a8	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.42, TFLite 2.17.0	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® X Elite	1.933 ms	39 - 39 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.509 ms	0 - 186 MB	NPU
RegNet	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.038 ms	0 - 44 MB	NPU
RegNet	ONNX	float	Qualcomm® QCS9075	3.075 ms	0 - 4 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.239 ms	0 - 140 MB	NPU
RegNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.021 ms	0 - 140 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® X Elite	1.137 ms	20 - 20 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.87 ms	0 - 197 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® QCS6490	27.939 ms	7 - 18 MB	CPU
RegNet	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.184 ms	0 - 107 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® QCS9075	1.324 ms	0 - 3 MB	NPU
RegNet	ONNX	w8a8	Qualcomm® QCM6690	18.049 ms	8 - 17 MB	CPU
RegNet	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.722 ms	0 - 152 MB	NPU
RegNet	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	13.384 ms	9 - 18 MB	CPU
RegNet	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.674 ms	0 - 153 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® X Elite	2.287 ms	1 - 1 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.413 ms	0 - 128 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	9.987 ms	1 - 76 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	2.051 ms	1 - 135 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8775P	3.263 ms	0 - 79 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® QCS9075	3.058 ms	1 - 3 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	3.482 ms	0 - 115 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA7255P	9.987 ms	1 - 76 MB	NPU
RegNet	QNN_DLC	float	Qualcomm® SA8295P	3.5 ms	1 - 66 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.106 ms	0 - 80 MB	NPU
RegNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.889 ms	1 - 81 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® X Elite	1.071 ms	0 - 0 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.638 ms	0 - 109 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS6490	2.712 ms	0 - 2 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	2.3 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	0.886 ms	0 - 2 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8775P	1.199 ms	0 - 78 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS9075	1.068 ms	0 - 2 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCM6690	7.075 ms	0 - 196 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	1.295 ms	0 - 110 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA7255P	2.3 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Qualcomm® SA8295P	1.606 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.493 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.153 ms	0 - 76 MB	NPU
RegNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.428 ms	0 - 80 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.417 ms	0 - 160 MB	NPU
RegNet	TFLITE	float	Qualcomm® QCS8275 (Proxy)	9.908 ms	0 - 104 MB	NPU
RegNet	TFLITE	float	Qualcomm® QCS8550 (Proxy)	2.031 ms	0 - 3 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8775P	13.985 ms	0 - 104 MB	NPU
RegNet	TFLITE	float	Qualcomm® QCS9075	3.046 ms	0 - 42 MB	NPU
RegNet	TFLITE	float	Qualcomm® QCS8450 (Proxy)	3.455 ms	0 - 144 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA7255P	9.908 ms	0 - 104 MB	NPU
RegNet	TFLITE	float	Qualcomm® SA8295P	3.52 ms	0 - 83 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.095 ms	0 - 106 MB	NPU
RegNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.887 ms	0 - 106 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.536 ms	0 - 119 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS6490	2.321 ms	0 - 22 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	1.992 ms	0 - 76 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.722 ms	0 - 2 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8775P	1.047 ms	0 - 77 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS9075	0.888 ms	0 - 22 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCM6690	6.625 ms	0 - 190 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	1.134 ms	0 - 119 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA7255P	1.992 ms	0 - 76 MB	NPU
RegNet	TFLITE	w8a8	Qualcomm® SA8295P	1.45 ms	0 - 71 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.428 ms	0 - 81 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	0.993 ms	0 - 72 MB	NPU
RegNet	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.393 ms	0 - 80 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.