How I get nodes names for input and output in Keras frozen model (.PB)

[vinicius1371]

Hello everybody,

I'm using a labview 2018 x64, and I trained a model with keras and tensorflow 1.4. I'm using IMAQ DL Model Run block, but the name of outout node that I put in this not work. One error appears talking that node name not included in graph. In my python code, the keras return predictions/Softmax with output node name but is not work in that block.

P.S. I'm using th aplication for image classification.

Thanks for yours help

[shindeiru]

I have the same problem! More specifically with YOLOv4 model. If anyone knows the solution, I would highly appreciate it.

[vinicius1371]

Hi, How are you?

My solition was use a python node block calling the python code. There is the possibility of frozen model but I think that spend more time.

[vinicius1371]

I can help you my email is viniciusdarochamotta@gmail.com.

[coconut456]

Hey, thanks for your reply. I have managed to list the output node names using the following script. Feel free to use it 🙂

The original script I modified is available on this link: https://stackoverflow.com/questions/47267636/tensorflow-how-do-i-find-my-output-node-in-my-tensorflo...

[vinicius1371]

I had a similar code, but my problem was a little bit different. I was using keras and I don't obtained success for to get a node by your names.

[shindeiru]

In the end I didn't manage to do it either. The names printed are not the node names, but operation tensor names which cannot be found using LabVIEW.

[vinicius1371]

I advise you to use the python node block to do this integration with that you create your code in python, taking the data from labview and passing it to your model to execute and then returning the processed data to labview, I got it like this.

[shindeiru]

Does it mean to execute the deep learning inference in a python script instead of the DL VI?

[vinicius1371]

Yes, if you want I can help you with that.