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 1.25.0	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:

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

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
StereoNet	ONNX	float	Snapdragon® X2 Elite	206.313 ms	5 - 5 MB	NPU
StereoNet	ONNX	float	Snapdragon® X Elite	375.632 ms	44 - 44 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	297.99 ms	6 - 4393 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	444.646 ms	0 - 49 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	199.12 ms	3 - 3301 MB	NPU
StereoNet	ONNX	float	Snapdragon® 8 Elite Mobile	251.742 ms	3 - 3235 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ Q-8750	251.742 ms	3 - 3235 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	375.632 ms	44 - 44 MB	NPU
StereoNet	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	530.939 ms	3 - 48 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® X2 Elite	193.3 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® X Elite	363.187 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	285.647 ms	3 - 4451 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® QCS8275	1294.016 ms	1 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	406.332 ms	3 - 6 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8775P	462.036 ms	2 - 3261 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8650P	462.036 ms	2 - 3261 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8255P	462.036 ms	2 - 3261 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	187.292 ms	3 - 3301 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA7255P	1294.016 ms	1 - 3260 MB	NPU
StereoNet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	237.606 ms	1 - 3245 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® SA8295P	515.862 ms	0 - 3367 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	237.606 ms	1 - 3245 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	363.187 ms	3 - 3 MB	NPU
StereoNet	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	511.428 ms	5 - 11 MB	NPU
StereoNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	401.979 ms	72 - 5322 MB	NPU
StereoNet	TFLITE	float	Qualcomm® SA8775P	5701.173 ms	2 - 33 MB	CPU
StereoNet	TFLITE	float	Qualcomm® SA8650P	5701.173 ms	2 - 33 MB	CPU
StereoNet	TFLITE	float	Qualcomm® SA8255P	5701.173 ms	2 - 33 MB	CPU
StereoNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	270.158 ms	72 - 3865 MB	NPU
StereoNet	TFLITE	float	Snapdragon® 8 Elite Mobile	275.619 ms	73 - 3773 MB	NPU
StereoNet	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	275.619 ms	73 - 3773 MB	NPU
StereoNet	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	661.282 ms	72 - 202 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.