openvino help needed

[owaisali]

i am using openvino in windows and got stucked after getting IR files of my unet model i dont know what to do in next step , please let me know if anyone could help me.

using Neural compute stick 2 

openvino 2021

[Adli]

Hi owaisali,

Thank you for reaching out to us. If possible, could you share additional information regarding the unet model? Is it a self-trained model or an intel model?

Since you already got the IR files, maybe you could run the benchmark app demo to check the model performance on Neural Compute Stick 2. Please refer to the following link:

https://docs.openvinotoolkit.org/latest/openvino_inference_engine_samples_benchmark_app_README.html#run_the_tool

Regards,

Adli

[owaisali]

I am using intel provided unet repository and after getting IR files ,i need your help how to run inference over NCS 2

[Adli]

Hi owaisali,

Thank you for the prompt response. Are you referring to this GitHub page ? A quick check, are you using a 2D U-Net model or a 3D U-Net model?

Regards,

Adli

[owaisali]

Thanks for your reply
I am using 2d model i have built it in VM using ubuntu and got IR files but help needed in how to test it over NCS 2 .your help regarding my issue will be much appreciated .
Thanks in advance
Engr Owais Ali

[Adli]

Hi owaisali,

Have you tried running the script 'plot_openvino_inference_examples.py' as mentioned on point 9 of the GitHub page? The example command:

python plot_openvino_inference_examples.py --device CPU

To infer on NCS 2 stick, please run the following example:

python plot_openvino_inference_examples.py --device MYRIAD

Please note to activate the OpenVINO environment before running the example command. To activate the environment, please run the following command:

source /opt/intel/openvino_2021/bin/setupvars.sh

Regards,

Adli

[owaisali]

I am also working over original unet code for that do i have to write my own inference file ?

[Adli]

Hi owaisali,

Have to say it really depends on what do you want to achieve in the end. I suggest that you try to infer your model with the original 'plot_openvino_inference_examples.py' script first. Feel free to post a question if you need help.

Regards,

Adli

[owaisali]

Much Obliged to have your reply, actually my aim is to train unet (2D) provided by intel on my dataset and the test it using Myrid NCS 2. In bold words i am really confused in word  "inference" meaning here because in the documentation of intel provided unet code they have run inference before converting the files for openvino.

If we can perform inference without IR files then what is the need of IR files here.

Refernce :

https://github.com/IntelAI/unet/tree/master/2D

Regards 

[Adli]

Hi owaisali,

For this case, you can perform inference without the IR files, but it is only limited to the Tensorflow framework. To perform inference without IR files, please run the 'plot_tf_inference_examples.py' script. IR files are required to perform inference using the OpenVINO program.

In your case, you want to test your self-trained UNet model using NCS2. The trained model must be converted to IR files first using OpenVINO's Model Optimizer and then load the IR files to the NCS2. Please note that you can't perform inference on NCS2 without the OpenVINO toolkit and the IR files. Please refer to this link: https://docs.openvinotoolkit.org/latest/index.html#packaging_and_deployment

Regards,

Adli

[Adli]

Hi owaisali,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Regards,

Adli