J721EXSOMXEVM: Vision Sdk TIDL: Steps for creating/training our own model and then running on J7 platform.

Hi,

I am new to this Machine learning world. I am learning new stuffs everyday but want to test my learning in parallel. I have trained and created basic mnist dataset on google colab and test its functionality. But now i want to integrate this model with J7 platform and create something like image classification vision app, so that in place of things it will detect numbers.

to give you the details....

my model is the simplest model having 2 dense hidden layer and expecting (28,28) tensors.

• can anyone help me with the steps of integrating my_model.pb with the platform - tried running the tidlmodelimport, but while running prints are coming till 'iterating 1 of 10000...' but nothing coming. So, i need to enable debug prints also.
• Also is there any google colab like cloud things where we can eliminate environment dependencies.