importing a ResNet model of python code in matlab?

I would like to load a deep network from the attached python code file and use it in the trainNetwork function in matlab.

Attached file: ResNet_simple_model.py, which contains the deep neural network model. (import tensorflow, numpy, sys, etc.)

How can I load and run this in matlab, load it as lgraph or layers, and train it with the trainNetwork(training_data, layers, options) function?

Waiting for help from experts.

[Constantino Carlos Reyes-Aldasoro]

Have you tried using the native ResNet CNNs (https://uk.mathworks.com/help/deeplearning/ref/resnet50.html)?

[Yongwon Jang]

Yes. I tried it with matlab toolbox.

But I want to know how to import and use .py file.

Thanks.

Answer 1

[Angelo Yeo]

1) Like @Constantino Carlos Reyes-Aldasoro said, you may apply to use the shipped ResNet for your research.

https://kr.mathworks.com/help/deeplearning/ref/resnet50.html

However, your Python model shows that your model is different from known ResNet models but rarther customized ones.

2) If you are using MATLAB R2021a or above, consider using importTensorFlowNetwork function. In order to import TF Network, you can save the model in "pb" format. How to save a TF Network into "pb" can be found in the link below.

https://www.tensorflow.org/guide/saved_model

[Yongwon Jang]

1) Yes. correct. It is customized one.

2) I want to know how to import and use .py file.

I'll study on this method.

Thank you.

[Yongwon Jang]

But still I wonder how to import and use .py file, for cowork with python user, not using importTensorFlowNetwork .

for example, the combinations of below...

pyversion

py.importlib.import_module

pyrunfile

:

:

:

[Angelo Yeo]

Can you describe your workflow? Why do you want to run "py" files via MATLAB?

In my imagination, I assumed that you might want to import trained network from Python's end. If you want to cowork with Python users you can also export the trained model with exportNetworkToTensorFlow.