Midas-V2: Optimized for Qualcomm Devices
Midas is designed for estimating depth at each point in an image.
This is based on the implementation of Midas-V2 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 |
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit Midas-V2 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 Midas-V2 on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.depth_estimation
Model Stats:
- Model checkpoint: MiDaS_small
- Input resolution: 256x256
- Number of parameters: 16.6M
- Model size (float): 63.2 MB
- Model size (w8a8): 16.9 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| Midas-V2 | ONNX | float | Snapdragon® X2 Elite | 1.311 ms | 2 - 2 MB | NPU |
| Midas-V2 | ONNX | float | Snapdragon® X Elite | 2.767 ms | 35 - 35 MB | NPU |
| Midas-V2 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.934 ms | 0 - 196 MB | NPU |
| Midas-V2 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 6.48 ms | 1 - 198 MB | NPU |
| Midas-V2 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.727 ms | 0 - 55 MB | NPU |
| Midas-V2 | ONNX | float | Qualcomm® QCS8450 | 6.48 ms | 1 - 198 MB | NPU |
| Midas-V2 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.257 ms | 0 - 163 MB | NPU |
| Midas-V2 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 4.042 ms | 1 - 47 MB | NPU |
| Midas-V2 | ONNX | float | Snapdragon® 8 Elite Mobile | 1.471 ms | 0 - 160 MB | NPU |
| Midas-V2 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.471 ms | 0 - 160 MB | NPU |
| Midas-V2 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.767 ms | 35 - 35 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® X2 Elite | 0.616 ms | 1 - 1 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® X Elite | 1.316 ms | 18 - 18 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.915 ms | 0 - 97 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.823 ms | 0 - 93 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.162 ms | 0 - 3 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.307 ms | 0 - 3 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® QCS8450 | 1.823 ms | 0 - 93 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.711 ms | 0 - 59 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.588 ms | 0 - 184 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.648 ms | 0 - 61 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.463 ms | 0 - 45 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 8.621 ms | 0 - 186 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.588 ms | 0 - 184 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.711 ms | 0 - 59 MB | NPU |
| Midas-V2 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.316 ms | 18 - 18 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® X2 Elite | 1.554 ms | 1 - 1 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® X Elite | 3.248 ms | 1 - 1 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.111 ms | 36 - 229 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 6.808 ms | 0 - 191 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® QCS8275 | 12.108 ms | 1 - 158 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.025 ms | 1 - 129 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® SA8775P | 4.254 ms | 1 - 159 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® SA8650P | 4.254 ms | 1 - 159 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® SA8255P | 4.254 ms | 1 - 159 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® QCS8450 | 6.808 ms | 0 - 191 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.296 ms | 1 - 163 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 4.46 ms | 1 - 3 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® SA7255P | 12.108 ms | 1 - 158 MB | NPU |
| Midas-V2 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.601 ms | 1 - 161 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® SA8295P | 5.451 ms | 1 - 155 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.601 ms | 1 - 161 MB | NPU |
| Midas-V2 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 3.248 ms | 1 - 1 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.724 ms | 0 - 0 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.451 ms | 0 - 0 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.911 ms | 0 - 85 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.87 ms | 0 - 83 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.534 ms | 2 - 4 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® QCS8275 | 2.881 ms | 0 - 47 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.317 ms | 0 - 2 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.569 ms | 0 - 50 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.569 ms | 0 - 50 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.569 ms | 0 - 50 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.87 ms | 0 - 83 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.661 ms | 0 - 46 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.577 ms | 0 - 172 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.582 ms | 0 - 50 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.412 ms | 0 - 2 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 8.945 ms | 0 - 174 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® SA7255P | 2.881 ms | 0 - 47 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 2.219 ms | 0 - 46 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.577 ms | 0 - 172 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.661 ms | 0 - 46 MB | NPU |
| Midas-V2 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.451 ms | 0 - 0 MB | NPU |
| Midas-V2 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.103 ms | 0 - 226 MB | NPU |
| Midas-V2 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 6.809 ms | 0 - 214 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® QCS8275 | 12.128 ms | 0 - 173 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 3.067 ms | 0 - 3 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® SA8775P | 4.275 ms | 0 - 173 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® SA8650P | 4.275 ms | 0 - 173 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® SA8255P | 4.275 ms | 0 - 173 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® QCS8450 | 6.809 ms | 0 - 214 MB | NPU |
| Midas-V2 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.295 ms | 0 - 175 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 4.017 ms | 0 - 39 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® SA7255P | 12.128 ms | 0 - 173 MB | NPU |
| Midas-V2 | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.607 ms | 0 - 175 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® SA8295P | 5.482 ms | 0 - 162 MB | NPU |
| Midas-V2 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.607 ms | 0 - 175 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.749 ms | 0 - 76 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.598 ms | 0 - 82 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 5.23 ms | 0 - 28 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® QCS8275 | 2.468 ms | 0 - 44 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.069 ms | 0 - 8 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® SA8775P | 1.343 ms | 0 - 49 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® SA8650P | 1.343 ms | 0 - 49 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® SA8255P | 1.343 ms | 0 - 49 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.598 ms | 0 - 82 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.557 ms | 0 - 42 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.339 ms | 0 - 171 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.494 ms | 0 - 47 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.19 ms | 0 - 20 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 8.265 ms | 0 - 171 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® SA7255P | 2.468 ms | 0 - 44 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® SA8295P | 1.957 ms | 0 - 43 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.339 ms | 0 - 171 MB | NPU |
| Midas-V2 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.557 ms | 0 - 42 MB | NPU |
License
- The license for the original implementation of Midas-V2 can be found here.
References
- Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer
- 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.
