library_name: pytorch
license: other
tags:
- android
pipeline_tag: depth-estimation
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:
- 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 | 206.507 ms | 5 - 5 MB | NPU |
| StereoNet | ONNX | float | Snapdragon® X Elite | 379.02 ms | 44 - 44 MB | NPU |
| StereoNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 299.657 ms | 7 - 4395 MB | NPU |
| StereoNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 414.696 ms | 0 - 49 MB | NPU |
| StereoNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 549.375 ms | 3 - 9 MB | NPU |
| StereoNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 379.02 ms | 44 - 44 MB | NPU |
| StereoNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 250.51 ms | 3 - 3256 MB | NPU |
| StereoNet | ONNX | float | Snapdragon® 8 Elite Mobile | 250.51 ms | 3 - 3256 MB | NPU |
| StereoNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 199.023 ms | 3 - 3299 MB | NPU |
| StereoNet | QNN_DLC | float | Snapdragon® X2 Elite | 192.888 ms | 3 - 3 MB | NPU |
| StereoNet | QNN_DLC | float | Snapdragon® X Elite | 366.296 ms | 3 - 3 MB | NPU |
| StereoNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 284.46 ms | 3 - 4455 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 1293.448 ms | 0 - 3262 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 473.426 ms | 3 - 286 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® SA8775P | 462.012 ms | 1 - 3262 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® SA8650P | 462.012 ms | 1 - 3262 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® SA8255P | 462.012 ms | 1 - 3262 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 450.22 ms | 5 - 11 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 366.296 ms | 3 - 3 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 236.415 ms | 0 - 3249 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® SA7255P | 1293.448 ms | 0 - 3262 MB | NPU |
| StereoNet | QNN_DLC | float | Qualcomm® SA8295P | 516.009 ms | 1 - 3347 MB | NPU |
| StereoNet | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 236.415 ms | 0 - 3249 MB | NPU |
| StereoNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 187.008 ms | 3 - 3303 MB | NPU |
| StereoNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 274.078 ms | 73 - 3772 MB | NPU |
| StereoNet | TFLITE | float | Snapdragon® 8 Elite Mobile | 274.078 ms | 73 - 3772 MB | NPU |
| StereoNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 278.603 ms | 73 - 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.