Depth-Anything-V2: Optimized for Mobile Deployment

Deep Convolutional Neural Network model for depth estimation

Depth Anything is designed for estimating depth at each point in an image.

This model is an implementation of Depth-Anything-V2 found here.

This repository provides scripts to run Depth-Anything-V2 on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

Model Type: Depth estimation
Model Stats:
- Model checkpoint: DepthAnything_V2_Small
- Input resolution: 518x518
- Number of parameters: 24.8M
- Model size: 94 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
Depth-Anything-V2	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	329.135 ms	0 - 86 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	378.932 ms	3 - 77 MB	FP16	NPU	Depth-Anything-V2.so
Depth-Anything-V2	Samsung Galaxy S23	Snapdragon® 8 Gen 2	ONNX	227.297 ms	0 - 63 MB	FP16	NPU	Depth-Anything-V2.onnx
Depth-Anything-V2	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	251.124 ms	0 - 248 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	287.78 ms	3 - 262 MB	FP16	NPU	Depth-Anything-V2.so
Depth-Anything-V2	Samsung Galaxy S24	Snapdragon® 8 Gen 3	ONNX	187.808 ms	0 - 986 MB	FP16	NPU	Depth-Anything-V2.onnx
Depth-Anything-V2	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	236.199 ms	1 - 271 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	240.939 ms	3 - 282 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	ONNX	156.392 ms	0 - 519 MB	FP16	NPU	Depth-Anything-V2.onnx
Depth-Anything-V2	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	329.447 ms	0 - 47 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	QCS8550 (Proxy)	QCS8550 Proxy	QNN	240.33 ms	4 - 5 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	SA7255P ADP	SA7255P	TFLITE	1138.667 ms	0 - 268 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	SA7255P ADP	SA7255P	QNN	1011.986 ms	2 - 12 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	SA8255 (Proxy)	SA8255P Proxy	TFLITE	331.644 ms	0 - 48 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	SA8255 (Proxy)	SA8255P Proxy	QNN	231.243 ms	4 - 5 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	SA8295P ADP	SA8295P	TFLITE	388.927 ms	1 - 274 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	SA8295P ADP	SA8295P	QNN	280.479 ms	6 - 12 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	SA8650 (Proxy)	SA8650P Proxy	TFLITE	329.235 ms	1 - 56 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	SA8650 (Proxy)	SA8650P Proxy	QNN	229.608 ms	4 - 5 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	SA8775P ADP	SA8775P	TFLITE	368.9 ms	1 - 269 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	SA8775P ADP	SA8775P	QNN	264.09 ms	3 - 13 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	361.044 ms	1 - 261 MB	FP16	NPU	Depth-Anything-V2.tflite
Depth-Anything-V2	QCS8450 (Proxy)	QCS8450 Proxy	QNN	439.422 ms	2 - 278 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	221.375 ms	3 - 3 MB	FP16	NPU	Use Export Script
Depth-Anything-V2	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	272.662 ms	62 - 62 MB	FP16	NPU	Depth-Anything-V2.onnx

Installation

This model can be installed as a Python package via pip.

pip install "qai-hub-models[depth_anything_v2]"

Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub with your Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.

With this API token, you can configure your client to run models on the cloud hosted devices.

qai-hub configure --api_token API_TOKEN

Navigate to docs for more information.

Demo off target

The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.

python -m qai_hub_models.models.depth_anything_v2.demo

The above demo runs a reference implementation of pre-processing, model inference, and post processing.

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.depth_anything_v2.demo

Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:

Performance check on-device on a cloud-hosted device
Downloads compiled assets that can be deployed on-device for Android.
Accuracy check between PyTorch and on-device outputs.

python -m qai_hub_models.models.depth_anything_v2.export

Profiling Results
------------------------------------------------------------
Depth-Anything-V2
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 329.1                  
Estimated peak memory usage (MB): [0, 86]                
Total # Ops                     : 635                    
Compute Unit(s)                 : NPU (635 ops)

How does this work?

This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:

Step 1: Compile model for on-device deployment

To compile a PyTorch model for on-device deployment, we first trace the model in memory using the jit.trace and then call the submit_compile_job API.

import torch

import qai_hub as hub
from qai_hub_models.models.depth_anything_v2 import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S23")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

Step 2: Performance profiling on cloud-hosted device

After compiling models from step 1. Models can be profiled model on-device using the target_model. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics.

profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)

Step 3: Verify on-device accuracy

To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.

input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.

Note: This on-device profiling and inference requires access to Qualcomm® AI Hub. Sign up for access.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.depth_anything_v2.demo --on-device

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.depth_anything_v2.demo -- --on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.
QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Depth-Anything-V2's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

The license for the original implementation of Depth-Anything-V2 can be found here.
The license for the compiled assets for on-device deployment can be found here

References

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.