I have trained a model using Microsoft's Custom Vision API. I can export the model in TensorFlow in the following formats: savedModel, TensorFlow Lite Float16, TensorFlow Lite, TensorFlow.js, TensorFlow. I can also export the model as a Dockerfile.
I want to download the model to perform transfer learning, but I have been unsuccessful in doing so. I believe I should use the savedModel format, but I'm unable to proceed. I have also tried converting the savedModel to .h5 format, which is more familiar to me, but I encountered issues with that as well.
When I use the command tf.saved_model.load(rutsavmod), I receive the following warning: "INFO:tensorflow:Saver not created because there are no variables in the graph to restore."
Has anyone ever performed transfer learning with a model exported from Custom Vision? Thank you.
Transfer learning with an exported Custom vision model
