HRNetFace: Optimized for Qualcomm Devices

Detects attributes (liveness, eye closeness, mask presence, glasses presence, sunglasses presence) that apply to a given face.

This is based on the implementation of HRNetFace 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 1.24.1	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.17.0	Download

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

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Model checkpoint: HR18-COFW.pth
• Input resolution: 256x256
• Number of parameters: 9.68M
• Model size (float): 36.87MB
• Model size (w8a8): 17.7 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
HRNetFace	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.327 ms	1 - 98 MB	NPU
HRNetFace	ONNX	float	Snapdragon® X2 Elite	1.532 ms	31 - 31 MB	NPU
HRNetFace	ONNX	float	Snapdragon® X Elite	3.313 ms	30 - 30 MB	NPU
HRNetFace	ONNX	float	Snapdragon® 8 Gen 3 Mobile	2.157 ms	0 - 155 MB	NPU
HRNetFace	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.978 ms	0 - 33 MB	NPU
HRNetFace	ONNX	float	Qualcomm® QCS9075	4.752 ms	2 - 4 MB	NPU
HRNetFace	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.65 ms	0 - 95 MB	NPU
HRNetFace	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.585 ms	0 - 105 MB	NPU
HRNetFace	ONNX	w8a8	Snapdragon® X2 Elite	1.427 ms	15 - 15 MB	NPU
HRNetFace	ONNX	w8a8	Snapdragon® X Elite	1.547 ms	15 - 15 MB	NPU
HRNetFace	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.887 ms	0 - 154 MB	NPU
HRNetFace	ONNX	w8a8	Qualcomm® QCS6490	91.751 ms	18 - 37 MB	CPU
HRNetFace	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.3 ms	0 - 21 MB	NPU
HRNetFace	ONNX	w8a8	Qualcomm® QCS9075	1.544 ms	0 - 3 MB	NPU
HRNetFace	ONNX	w8a8	Qualcomm® QCM6690	49.856 ms	18 - 37 MB	CPU
HRNetFace	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.696 ms	0 - 95 MB	NPU
HRNetFace	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	46.773 ms	18 - 36 MB	CPU
HRNetFace	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.37 ms	1 - 83 MB	NPU
HRNetFace	QNN_DLC	float	Snapdragon® X2 Elite	1.955 ms	1 - 1 MB	NPU
HRNetFace	QNN_DLC	float	Snapdragon® X Elite	3.71 ms	1 - 1 MB	NPU
HRNetFace	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	2.318 ms	0 - 125 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	15.446 ms	1 - 77 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	3.326 ms	1 - 147 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® SA8775P	5.117 ms	1 - 80 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® QCS9075	5.028 ms	1 - 4 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	5.415 ms	0 - 109 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® SA7255P	15.446 ms	1 - 77 MB	NPU
HRNetFace	QNN_DLC	float	Qualcomm® SA8295P	5.619 ms	1 - 62 MB	NPU
HRNetFace	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.748 ms	0 - 79 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.528 ms	0 - 72 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.856 ms	0 - 0 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® X Elite	1.556 ms	0 - 0 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.918 ms	0 - 108 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCS6490	3.643 ms	0 - 2 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	3.238 ms	0 - 70 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.329 ms	0 - 2 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® SA8775P	1.702 ms	0 - 72 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCS9075	1.615 ms	0 - 2 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCM6690	10.463 ms	0 - 197 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	1.782 ms	0 - 107 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® SA7255P	3.238 ms	0 - 70 MB	NPU
HRNetFace	QNN_DLC	w8a8	Qualcomm® SA8295P	2.208 ms	0 - 68 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.673 ms	0 - 71 MB	NPU
HRNetFace	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.643 ms	0 - 78 MB	NPU
HRNetFace	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.383 ms	0 - 100 MB	NPU
HRNetFace	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	2.345 ms	0 - 143 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® QCS8275 (Proxy)	15.416 ms	0 - 93 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® QCS8550 (Proxy)	3.375 ms	0 - 2 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® SA8775P	5.162 ms	0 - 96 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® QCS9075	5.038 ms	0 - 35 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® QCS8450 (Proxy)	5.44 ms	0 - 129 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® SA7255P	15.416 ms	0 - 93 MB	NPU
HRNetFace	TFLITE	float	Qualcomm® SA8295P	5.652 ms	0 - 76 MB	NPU
HRNetFace	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.757 ms	0 - 89 MB	NPU
HRNetFace	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.466 ms	0 - 73 MB	NPU
HRNetFace	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.75 ms	0 - 117 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCS6490	3.319 ms	0 - 18 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	2.906 ms	0 - 70 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	1.097 ms	0 - 3 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® SA8775P	1.488 ms	0 - 73 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCS9075	1.325 ms	0 - 18 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCM6690	9.776 ms	0 - 197 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	1.49 ms	0 - 110 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® SA7255P	2.906 ms	0 - 70 MB	NPU
HRNetFace	TFLITE	w8a8	Qualcomm® SA8295P	1.961 ms	0 - 68 MB	NPU
HRNetFace	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.581 ms	0 - 71 MB	NPU
HRNetFace	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.362 ms	0 - 75 MB	NPU

License

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

References

• Deep High-Resolution Representation Learning for Visual 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.