StereoNet: Optimized for Qualcomm Devices

StereoNet is an end-to-end deep architecture for real-time stereo matching that produces high-quality, edge-preserved disparity maps from a rectified stereo image pair.

This is based on the implementation of StereoNet 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
TFLITE	float	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.depth_estimation

Model Stats:

• Input resolution: 786x490
• Model checkpoint: KeystoneDepth (epoch=21-step=696366.ckpt)
• Model size (float): 7.41 MB
• Number of parameters: 1.94M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
StereoNet	ONNX	float	Snapdragon® X2 Elite	208.109 ms	5 - 5 MB	NPU
StereoNet	ONNX	float	Snapdragon® X Elite	377.314 ms	44 - 44 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	299.963 ms	6 - 4400 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	492.311 ms	3 - 9 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	458.139 ms	4 - 8 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	464.167 ms	3 - 9 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	377.314 ms	44 - 44 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ Q-8750	249.731 ms	3 - 3236 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Elite Mobile	249.731 ms	3 - 3236 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	198.933 ms	3 - 3300 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® X2 Elite	191.479 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® X Elite	363.646 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	286.692 ms	3 - 4451 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	482.377 ms	3 - 10 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	1294.512 ms	2 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	463.723 ms	3 - 7 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8775P	462.148 ms	1 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8650P	462.148 ms	1 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8255P	462.148 ms	1 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	448.748 ms	3 - 9 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	363.646 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	235.592 ms	0 - 3243 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA7255P	1294.512 ms	2 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8295P	515.7 ms	0 - 3366 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	235.592 ms	0 - 3243 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	188.072 ms	3 - 3302 MB	NPU
StereoNet	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	302.03 ms	73 - 3773 MB	NPU
StereoNet	TFLITE	float	Snapdragon® 8 Elite Mobile	302.03 ms	73 - 3773 MB	NPU
StereoNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	274.392 ms	72 - 3863 MB	NPU

License

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

References

• StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth Prediction
• 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.