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	1.314 ms	212 - 212 MB	NPU
Midas-V2	ONNX	float	Snapdragon® X Elite	2.771 ms	148 - 148 MB	NPU
Midas-V2	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.924 ms	0 - 186 MB	NPU
Midas-V2	ONNX	float	Snapdragon® 8 Gen 1 Mobile	6.487 ms	1 - 186 MB	NPU
Midas-V2	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.726 ms	1 - 56 MB	NPU
Midas-V2	ONNX	float	Qualcomm® QCS8450	6.487 ms	1 - 186 MB	NPU
Midas-V2	ONNX	float	Snapdragon® 8 Elite Mobile	1.469 ms	0 - 160 MB	NPU
Midas-V2	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.243 ms	1 - 162 MB	NPU
Midas-V2	ONNX	float	Qualcomm® QCS9075	4.042 ms	1 - 47 MB	NPU
Midas-V2	ONNX	float	Qualcomm® QCS8750	1.469 ms	0 - 160 MB	NPU
Midas-V2	ONNX	float	Qualcomm® QCS7181	2.771 ms	148 - 148 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® X2 Elite	0.616 ms	213 - 213 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® X Elite	1.323 ms	149 - 149 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.907 ms	0 - 84 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	1.838 ms	0 - 84 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS6490	4.117 ms	0 - 45 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.326 ms	0 - 31 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS8450	1.838 ms	0 - 84 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.642 ms	0 - 55 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	1.599 ms	0 - 177 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCM6690	8.623 ms	0 - 181 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS9075	1.463 ms	0 - 45 MB	NPU
Midas-V2	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.715 ms	0 - 60 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS7790	1.599 ms	0 - 177 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS8750	0.715 ms	0 - 60 MB	NPU
Midas-V2	ONNX	w8a8	Qualcomm® QCS7181	1.323 ms	149 - 149 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® X2 Elite	1.58 ms	1 - 1 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® X Elite	3.31 ms	1 - 1 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	2.114 ms	0 - 191 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	6.825 ms	0 - 193 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS8275	12.08 ms	1 - 158 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	3.024 ms	1 - 3 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS8450	6.825 ms	0 - 193 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.6 ms	0 - 159 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® SA8295P	5.443 ms	1 - 155 MB	NPU
Midas-V2	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.284 ms	1 - 165 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® SA7255P	12.08 ms	1 - 158 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS9075	4.46 ms	1 - 3 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS8750	1.6 ms	0 - 159 MB	NPU
Midas-V2	QNN_DLC	float	Qualcomm® QCS7181	3.31 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	1.456 ms	0 - 0 MB	NPU
Midas-V2	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.92 ms	0 - 81 MB	NPU
Midas-V2	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.874 ms	0 - 84 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS6490	4.065 ms	2 - 4 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS8275	2.844 ms	0 - 47 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.321 ms	0 - 3 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS8450	1.874 ms	0 - 84 MB	NPU
Midas-V2	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.579 ms	0 - 50 MB	NPU
Midas-V2	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.572 ms	0 - 171 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCM6690	9.053 ms	2 - 176 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS9075	1.439 ms	0 - 2 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® SA7255P	2.844 ms	0 - 47 MB	NPU
Midas-V2	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.654 ms	0 - 46 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® SA8295P	2.24 ms	0 - 47 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS7790	1.572 ms	0 - 171 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS8750	0.654 ms	0 - 46 MB	NPU
Midas-V2	QNN_DLC	w8a8	Qualcomm® QCS7181	1.456 ms	0 - 0 MB	NPU
Midas-V2	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	2.109 ms	0 - 226 MB	NPU
Midas-V2	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	6.828 ms	0 - 215 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® QCS8275	12.14 ms	0 - 173 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® QCS8550 (Proxy)	3.014 ms	0 - 3 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® SA8775P	22.047 ms	0 - 32 MB	GPU
Midas-V2	TFLITE	float	Qualcomm® SA8650P	22.047 ms	0 - 32 MB	GPU
Midas-V2	TFLITE	float	Qualcomm® SA8255P	22.047 ms	0 - 32 MB	GPU
Midas-V2	TFLITE	float	Qualcomm® QCS8450	6.828 ms	0 - 215 MB	NPU
Midas-V2	TFLITE	float	Snapdragon® 8 Elite Mobile	1.606 ms	0 - 171 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® SA8295P	5.49 ms	0 - 162 MB	NPU
Midas-V2	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.282 ms	0 - 177 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® SA7255P	12.14 ms	0 - 173 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® QCS9075	4.391 ms	0 - 39 MB	NPU
Midas-V2	TFLITE	float	Qualcomm® QCS8750	1.606 ms	0 - 171 MB	NPU
Midas-V2	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.744 ms	0 - 77 MB	NPU
Midas-V2	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	1.595 ms	0 - 82 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS6490	3.647 ms	0 - 28 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS8275	2.46 ms	0 - 44 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	1.061 ms	0 - 4 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® SA8775P	22.388 ms	1 - 32 MB	GPU
Midas-V2	TFLITE	w8a8	Qualcomm® SA8650P	22.388 ms	1 - 32 MB	GPU
Midas-V2	TFLITE	w8a8	Qualcomm® SA8255P	22.388 ms	1 - 32 MB	GPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS8450	1.595 ms	0 - 82 MB	NPU
Midas-V2	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.493 ms	0 - 44 MB	NPU
Midas-V2	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.344 ms	0 - 168 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCM6690	8.211 ms	0 - 171 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS9075	1.174 ms	0 - 20 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® SA7255P	2.46 ms	0 - 44 MB	NPU
Midas-V2	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	0.554 ms	0 - 49 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® SA8295P	1.964 ms	0 - 43 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS7790	1.344 ms	0 - 168 MB	NPU
Midas-V2	TFLITE	w8a8	Qualcomm® QCS8750	0.554 ms	0 - 49 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.