IssueTracker

I didn't find the component for AutoML vision. Sorry if this falls (it seems that I typed "fails" three times already) into wrong bucket. This is the last ticket that I will create today :-). This is a great product and I sincerely want to see it becomes better.

This is not an urgent issue nor critical to what I need to do. I know the workarounds even if it does not get addressed.

I did not find any documentation on how to export the final model structure (i.e., how many layers, what are each layer) as well as parameter values. Is this what supposed to be (i.e., Google owns the TensorFlow model object, but only opens up a API for us to run predictions on it)?

I can certainly understand the commercial reason for this choice, but this severely limits our possibility of "commercialise" models trained by AutoML Vision. In the medical field, it will be extremely difficult to get the doctors to use something that even its developer/creator doesn't know what's in it.

Thank you so much.
Cheers,
Seagle