SINet: Optimized for Qualcomm Devices

SINet is a machine learning model that is designed to segment people from close-up portrait images in real time.

This is based on the implementation of SINet 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
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

Input resolution: 224x224
Model checkpoint: SINet.pth
Model size (float): 415 KB
Model size (w8a8): 241 KB
Number of output classes: 2 (foreground / background)
Number of parameters: 91.9K

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
SINet	ONNX	float	Snapdragon® X2 Elite	1.078 ms	2 - 2 MB	NPU
SINet	ONNX	float	Snapdragon® X Elite	1.827 ms	2 - 2 MB	NPU
SINet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.178 ms	0 - 49 MB	NPU
SINet	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.117 ms	1 - 57 MB	NPU
SINet	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	1.903 ms	1 - 5 MB	NPU
SINet	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.759 ms	0 - 3 MB	NPU
SINet	ONNX	float	Qualcomm® QCS8450	2.117 ms	1 - 57 MB	NPU
SINet	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	1.996 ms	1 - 5 MB	NPU
SINet	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.827 ms	2 - 2 MB	NPU
SINet	ONNX	float	Qualcomm® Dragonwing™ Q-8750	0.889 ms	0 - 35 MB	NPU
SINet	ONNX	float	Snapdragon® 8 Elite Mobile	0.889 ms	0 - 35 MB	NPU
SINet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.748 ms	0 - 40 MB	NPU
SINet	QNN_DLC	float	Snapdragon® X2 Elite	0.971 ms	1 - 1 MB	NPU
SINet	QNN_DLC	float	Snapdragon® X Elite	1.982 ms	1 - 1 MB	NPU
SINet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.165 ms	0 - 45 MB	NPU
SINet	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	2.05 ms	0 - 53 MB	NPU
SINet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	1.824 ms	1 - 4 MB	NPU
SINet	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.715 ms	1 - 2 MB	NPU
SINet	QNN_DLC	float	Qualcomm® SA8775P	2.191 ms	0 - 36 MB	NPU
SINet	QNN_DLC	float	Qualcomm® SA8650P	2.191 ms	0 - 36 MB	NPU
SINet	QNN_DLC	float	Qualcomm® SA8255P	2.191 ms	0 - 36 MB	NPU
SINet	QNN_DLC	float	Qualcomm® QCS8450	2.05 ms	0 - 53 MB	NPU
SINet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	1.987 ms	3 - 6 MB	NPU
SINet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	1.982 ms	1 - 1 MB	NPU
SINet	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	0.867 ms	0 - 31 MB	NPU
SINet	QNN_DLC	float	Qualcomm® SA7255P	3.927 ms	1 - 33 MB	NPU
SINet	QNN_DLC	float	Qualcomm® SA8295P	2.603 ms	0 - 34 MB	NPU
SINet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	0.867 ms	0 - 31 MB	NPU
SINet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.727 ms	1 - 41 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® X2 Elite	1.1 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® X Elite	2.196 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	1.296 ms	0 - 49 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	2.471 ms	0 - 55 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	1.823 ms	0 - 3 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.898 ms	0 - 35 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® SA8775P	2.235 ms	0 - 41 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® SA8650P	2.235 ms	0 - 41 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® SA8255P	2.235 ms	0 - 41 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® QCS8450	2.471 ms	0 - 55 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	2.074 ms	0 - 3 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	2.196 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	0.979 ms	0 - 37 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® SA7255P	3.692 ms	0 - 39 MB	NPU
SINet	QNN_DLC	w8a16	Qualcomm® SA8295P	3.07 ms	0 - 37 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	0.979 ms	0 - 37 MB	NPU
SINet	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.797 ms	0 - 40 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.774 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® X Elite	1.603 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.969 ms	0 - 47 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.666 ms	0 - 52 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	1.388 ms	0 - 3 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.396 ms	0 - 2 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® SA8775P	1.677 ms	0 - 39 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® SA8650P	1.677 ms	0 - 39 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® SA8255P	1.677 ms	0 - 39 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® QCS8450	1.666 ms	0 - 52 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.535 ms	2 - 4 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	1.603 ms	0 - 0 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	0.722 ms	0 - 41 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® SA7255P	2.794 ms	0 - 36 MB	NPU
SINet	QNN_DLC	w8a8	Qualcomm® SA8295P	2.169 ms	0 - 36 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.722 ms	0 - 41 MB	NPU
SINet	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.567 ms	0 - 37 MB	NPU
SINet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.212 ms	0 - 43 MB	NPU
SINet	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2.282 ms	0 - 54 MB	NPU
SINet	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	2.225 ms	0 - 6 MB	NPU
SINet	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.82 ms	0 - 3 MB	NPU
SINet	TFLITE	float	Qualcomm® SA8775P	2.628 ms	0 - 36 MB	NPU
SINet	TFLITE	float	Qualcomm® SA8650P	2.628 ms	0 - 36 MB	NPU
SINet	TFLITE	float	Qualcomm® SA8255P	2.628 ms	0 - 36 MB	NPU
SINet	TFLITE	float	Qualcomm® QCS8450	2.282 ms	0 - 54 MB	NPU
SINet	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2.297 ms	2 - 6 MB	NPU
SINet	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	0.935 ms	0 - 35 MB	NPU
SINet	TFLITE	float	Qualcomm® SA7255P	4.499 ms	1 - 34 MB	NPU
SINet	TFLITE	float	Qualcomm® SA8295P	3.084 ms	0 - 33 MB	NPU
SINet	TFLITE	float	Snapdragon® 8 Elite Mobile	0.935 ms	0 - 35 MB	NPU
SINet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.8 ms	0 - 33 MB	NPU

License

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

References

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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/SINet

SINet: Extreme Lightweight Portrait Segmentation Networks with Spatial Squeeze Modules and Information Blocking Decoder

Paper • 1911.09099 • Published Nov 20, 2019