Depth-Anything-V3: Optimized for Qualcomm Devices

Depth Anything 3 (DA3), a model that predicts spatially consistent geometry from arbitrary visual inputs, with or without known camera poses.

This is based on the implementation of Depth-Anything-V3 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 Depth-Anything-V3 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 Depth-Anything-V3 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.depth_estimation

Model Stats:

• Model checkpoint: da3-small
• Input resolution: 518x518
• Number of parameters: 24.7M
• Model size (float): 94.3 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Depth-Anything-V3	ONNX	float	Snapdragon® X2 Elite	27.979 ms	210 - 210 MB	NPU
Depth-Anything-V3	ONNX	float	Snapdragon® X Elite	70.364 ms	179 - 179 MB	NPU
Depth-Anything-V3	ONNX	float	Snapdragon® 8 Gen 3 Mobile	55.057 ms	5 - 646 MB	NPU
Depth-Anything-V3	ONNX	float	Snapdragon® 8 Gen 1 Mobile	129.194 ms	5 - 654 MB	NPU
Depth-Anything-V3	ONNX	float	Qualcomm® QCS8550 (Proxy)	69.21 ms	0 - 81 MB	NPU
Depth-Anything-V3	ONNX	float	Qualcomm® QCS8450	129.194 ms	5 - 654 MB	NPU
Depth-Anything-V3	ONNX	float	Snapdragon® 8 Elite Mobile	36.649 ms	2 - 500 MB	NPU
Depth-Anything-V3	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	28.29 ms	3 - 518 MB	NPU
Depth-Anything-V3	ONNX	float	Qualcomm® QCS9075	81.427 ms	4 - 49 MB	NPU
Depth-Anything-V3	ONNX	float	Qualcomm® QCS8750	36.649 ms	2 - 500 MB	NPU
Depth-Anything-V3	ONNX	float	Qualcomm® QCS7181	70.364 ms	179 - 179 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® X2 Elite	35.178 ms	3 - 3 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® X Elite	78.942 ms	3 - 3 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	61.065 ms	0 - 644 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	150.506 ms	0 - 675 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS8275	186.786 ms	2 - 492 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	80.643 ms	3 - 417 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS8450	150.506 ms	0 - 675 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® 8 Elite Mobile	43.145 ms	3 - 504 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® SA8295P	118.767 ms	0 - 504 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	33.344 ms	2 - 520 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® SA7255P	186.786 ms	2 - 492 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS9075	98.235 ms	5 - 11 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS8750	43.145 ms	3 - 504 MB	NPU
Depth-Anything-V3	QNN_DLC	float	Qualcomm® QCS7181	78.942 ms	3 - 3 MB	NPU
Depth-Anything-V3	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	54.351 ms	1 - 596 MB	NPU
Depth-Anything-V3	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	118.325 ms	1 - 611 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® QCS8275	180.792 ms	1 - 467 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® QCS8550 (Proxy)	74.322 ms	1 - 9 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® SA8775P	935.6 ms	25 - 50 MB	CPU
Depth-Anything-V3	TFLITE	float	Qualcomm® SA8650P	935.6 ms	25 - 50 MB	CPU
Depth-Anything-V3	TFLITE	float	Qualcomm® SA8255P	935.6 ms	25 - 50 MB	CPU
Depth-Anything-V3	TFLITE	float	Qualcomm® QCS8450	118.325 ms	1 - 611 MB	NPU
Depth-Anything-V3	TFLITE	float	Snapdragon® 8 Elite Mobile	39.229 ms	1 - 474 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® SA8295P	103.452 ms	1 - 466 MB	NPU
Depth-Anything-V3	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	30.711 ms	1 - 488 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® SA7255P	180.792 ms	1 - 467 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® QCS9075	95.467 ms	0 - 80 MB	NPU
Depth-Anything-V3	TFLITE	float	Qualcomm® QCS8750	39.229 ms	1 - 474 MB	NPU

License

• The license for the original implementation of Depth-Anything-V3 can be found here.

References

• Depth Anything 3: Recovering the visual space from any views
• 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.