Can I import a Tensor Flow .pb model to Perceptilabs

Hello, I am testing Perceptilabs using the Free version. My question is am I able to import a Tensor Flow .pb model? I tried doing an import but Perceptilabs doesn’t seem to see the file. I am not sure if this is because I am using the free version or that is just not a capability. Under Export there is an option for saving as a Tensor Flow model so I figured it would let me import one as well.

Hi @vmcharca,
At the moment you can only import PerceptiLabs models into PerceptiLabs, but we have a new feature in the pipeline which will do just that. With that feature you will be able to load any TensorFlow .pb file and get it visualized in PerceptiLabs interface.

All the best,

1 Like

That would be a very nice feature to have that way we can convert existing models we have. Any thoughts on when that might be available?

I’m glad you are liking the idea of the feature!
It’s already come a bit in development, but we are currently focusing our efforts on making the data processing part a bit better, as well as adding common pre-trained models as components (such as VGG16 etc.).
After that is implemented we will re-focus on the feature to import a TensorFlow pb file. I expect a first version to be in early Q2 this year.

If you don’t mind me asking, what would be the biggest benefit you get out of importing you existing models into PerceptiLabs?
Is it to easier manage them with a visual interface, or that you can easier compare them to any new models you may create?

To be able to see them visually, it may help to review them and improve their performance

That makes sense, thank you for letting me know :slight_smile:

Question what are the limitations I have using the free version?

Hey again @vmcharca,
Right now, the only limitations are that you are unable to run the training distributed and that everything has to be ran locally.
If you are interested in an enterprise version, as well as discussing our feature roadmap, I would be more than happy to set up a call between us.