Running a Tensorflow model in NI Vision Dev. Module 2018

[cfosalau]

I am trying to test a signal processing model trained with Tensorflow using the Deep Learning from NI Vision Development Module 2018. I loaded the model with IMAQ DL Model Create function and now I want to use it with an array dataset with IMAQ DL Model Run, but I do not succeed to find the “Output Model Names” of the model required by the function. I tried to find it with model,input in Tensorflow, but the item it returns does not fit the Labview function. Can anybody help me? Thank you.

[BeeRio · Accepted solution]

Firstly, you need to to freeze your model weights in .pb format as described here

To get the model input output names, run the following commands after you define the model:

print('inputs: ', [input.op.name for input in model.inputs])

print('outputs: ', [output.op.name for output in model.outputs])

PS. I use above commands when defining my model in Keras.

[cfosalau]

great! It works! Thank you very much, BeeRio.

[BeeRio]

Glad it worked. You can mark that as solution.

Thanks.