Midas-V2-Quantized: Optimized for Mobile Deployment

Quantized Deep Convolutional Neural Network model for depth estimation

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

This model is an implementation of Midas-V2-Quantized found here.

This repository provides scripts to run Midas-V2-Quantized 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: MiDaS_small
- Input resolution: 256x256
- Number of parameters: 16.6M
- Model size: 16.6 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
Midas-V2-Quantized	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	1.093 ms	0 - 1 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	1.436 ms	0 - 7 MB	INT8	NPU	Midas-V2-Quantized.so
Midas-V2-Quantized	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	0.773 ms	0 - 89 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	1.013 ms	0 - 25 MB	INT8	NPU	Midas-V2-Quantized.so
Midas-V2-Quantized	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	0.721 ms	0 - 47 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	0.984 ms	0 - 21 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	RB3 Gen 2 (Proxy)	QCS6490 Proxy	TFLITE	3.67 ms	0 - 48 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	RB3 Gen 2 (Proxy)	QCS6490 Proxy	QNN	6.075 ms	0 - 8 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	RB5 (Proxy)	QCS8250 Proxy	TFLITE	15.347 ms	0 - 3 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	1.083 ms	0 - 1 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	QCS8550 (Proxy)	QCS8550 Proxy	QNN	1.308 ms	0 - 1 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	SA8255 (Proxy)	SA8255P Proxy	TFLITE	1.089 ms	0 - 12 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	SA8255 (Proxy)	SA8255P Proxy	QNN	1.318 ms	0 - 1 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	SA8775 (Proxy)	SA8775P Proxy	TFLITE	1.101 ms	0 - 1 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	SA8775 (Proxy)	SA8775P Proxy	QNN	1.312 ms	0 - 2 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	SA8650 (Proxy)	SA8650P Proxy	TFLITE	1.078 ms	0 - 16 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	SA8650 (Proxy)	SA8650P Proxy	QNN	1.32 ms	0 - 2 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	SA8295P ADP	SA8295P	TFLITE	1.928 ms	0 - 46 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	SA8295P ADP	SA8295P	QNN	2.526 ms	0 - 6 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	1.417 ms	0 - 87 MB	INT8	NPU	Midas-V2-Quantized.tflite
Midas-V2-Quantized	QCS8450 (Proxy)	QCS8450 Proxy	QNN	1.776 ms	0 - 25 MB	INT8	NPU	Use Export Script
Midas-V2-Quantized	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	1.479 ms	0 - 0 MB	INT8	NPU	Use Export Script

Installation

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

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

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.midas_quantized.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.midas_quantized.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.midas_quantized.export

Profiling Results
------------------------------------------------------------
Midas-V2-Quantized
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 1.1                    
Estimated peak memory usage (MB): [0, 1]                 
Total # Ops                     : 145                    
Compute Unit(s)                 : NPU (145 ops)

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.midas_quantized.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.midas_quantized.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 Midas-V2-Quantized's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

The license for the original implementation of Midas-V2-Quantized 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.