Support for installing SKlearn and TensorFlow libraries on i.MX8M

[Sunil77]

I am trying to develop Machine Learning (ML) applications on the i.MX 8M (Yocto build procedure). Accordingly, I would like to have SKlearn and Tensorflow libraries installed in our Linux environment. I came to know that iMX 8M could be used only for deployment (only trained Tensorflow lite models could be loaded). Could we do training and modelling in iMX 8M (developmental activities for ML applications). Could you please share any inputs (patches/ installation procedures) on the same.

[brian14]

Hi @Sunil77,

In our BSP, TensorFlow Lite is included instead of TensorFlow because of this library contains many packages unnecessary to deploy your machine learning model.
For embedded systems with a constrained memory, you will need to develop your machine learning model using a host PC and then export to a TensorFlow Lite model.

Finally, in the i.MX8M Plus you can use Python or C++ API for TFLite interpreter and obtain inferences.