Completely lost on beginner instructions

[onthegrid007]

Hello,

So to begin I am working on a project that compiles and trains a .blob file for use in a Luxonis Oak-D DepthAI camera for neuralnetwork processing. I have all of the code to run the camera with the blob file and use the neural network output in processing, my errors start in the building of the dataset and blob file. I have just completed a program that generates a folder structure as follows:

dataset/

    class1/

        324.jpg

        512.jpg

    class2/

        128.jpg

        321.jpg

Where dataset is the name of the parent dataset folder, class1 and class 2 are the lables of the data to be trained upon. This dataset is then taken at random and split into training and validation datasets to be used for training.

I have done some research to create and train a tensorflow model and the convert to an onyx model using the following command:
python.exe -m tf2onnx.convert --saved-model saved_model --output "${ONNX_MODEL}"

Then the model optimizer command works:

mo --input_model "${ONNX_MODEL}" --output_dir "${IR_MODEL_DIR}"

Then Converting IR model to .blob format is where I am having the trouble, I have no instalation location for the myriad_convert and the mo flags to do this do not seem to work:

mo --input_model "${IR_MODEL_DIR}/model.xml" --data_type FP16 --compile_tool_config --output_dir "${IR_MODEL_DIR}" --myriad_target="${BLOB_FILE}"
This fails with unknown arguments: --data_type FP16 --compile_tool_config --myriad_target="${BLOB_FILE}"

Any help would be great on what I am doing right and what I am doing wrong. Thank you.

-Peter F

[Wan_Intel]

Hi onthegrid007,

Thanks for reaching out to us.

Could you please share the following information with us to further investigate the issue?

• If Python API, version used
• OpenVINO™ toolkit version used
• Hardware specifications
• Host Operating System

Regards,

Wan

[onthegrid007]

Python 10

tensorflow 2.10

tf2onnx 1.13.0

openvino-dev[onnx,tensorflow] 2023.0

The blob file compile hardware is an x64 intel processor with a Nividia 3060 GPU

The camera running the blob file is a Luxonis DepthAI OAK-D unit

Host OS is Windows 11 x64

[Wan_Intel]

Hi onthegrid007,

Thanks for the information.

I'm glad to know that you had successfully converted your model into OpenVINO™ Intermediate Representation format (IR). To convert IR model into a .blob file, you need to utilize OpenVINO™ Compile Tool.

The OpenVINO™ Compile Tool is available at OpenVINO™ toolkit 2022.3.2. Please download OpenVINO™ toolkit 2022.3.2 archive file from the following link:

https://storage.openvinotoolkit.org/repositories/openvino/packages/2022.3.2/windows/w_openvino_toolkit_windows_2022.3.2.9279.e2c7e4d7b4d_x86_64.zip

After extracted the .zip file, OpenVINO™ Compile Tool can be found at the following directory:

<OpenVINO™_install_dir>\tools\compile_tool

You may use the following command to convert your IR model into a .blob file:

compile_tool.exe -m <path_to_IR>\model_name.xml -d CPU

Regards,

Wan

[onthegrid007]

This is where I fail to understand what to do, I have downloaded this exact zip file like 5 times (I just did it again from the link you sent me) and the directory structure is as follows:

version.txt

bin/

include/

cmake

lib/

3rdparty/

There is no tools/ directory at all and for that matter absolutly NO .exe files anywhere in the .zip file

[Wan_Intel]

Hi onthegrid007,

Thanks for the information.

After you've extracted the .zip file, you may go to the following directory:

C:\Users\username\Downloads\w_openvino_toolkit_windows_2022.3.2.9279.e2c7e4d7b4d_x86_64\w_openvino_toolkit_windows_2022.3.2.9279.e2c7e4d7b4d_x86_64\tools\compile_tool

Regards,

Wan

[Wan_Intel]

Hi onthegrid007,

Thanks for the question.

If you need additional information from Intel, please submit a new question as this thread will no longer be monitored.

Regards,

Wan