Re: Putting Keras ML model inside Kivy

[Robert]

You are jumping in at the deep end of the swimming pool, I hope you can.....

Keras and Tensorfow form a development environment, you can develop with this on the desktop. Then convert the model to the format for the Tensorflow-lite run time,  this runtime package you can use on mobile (or desktop) to evaluate your converted model.

Here is a classification example using Kivy and tflite.

https://github.com/Android-for-Python/c4k_tflite_example

The README contains a short implementation overview, test results, and links to some tflite documentation.

Particularly https://github.com/tensorflow/examples/tree/master/lite/examples  (not all of which are video.)

The data source in the example is video frames provided by the Preview widget

https://github.com/Android-for-Python/camera4kivy#image-analysis

Details of working with TF models are outside the scope of this group.

On Friday, November 18, 2022 at 8:07:25 AM UTC-10 fran...@gmail.com wrote:

Hi All:

This is my first post to the group.

I am working on a project where I want to put my Keras ML model (model.h5,) into Kivy, to make a mobile machine learning app.

I have looked online, on Youtube and at Amazon.com for info and books that would walk you through the process.  But I could not find a book that discussed exactly how to do this.  I did find some Youtube videos that were helpful.

If you know of a book or if you know how to do this, I would appreciate some help.

I have my Kivy app built, I have put my model.h5 file inside the app.  Now, I need to connect the CNN model to a button to acctivate the model.

Note:

This is a classification app 

This is my first Kivy project.

Any help, suggestions or book recommendation would be most appreciated.

Cheers!

Frank