K Means Clustering Integration to Power Platform

K Means Clustering Integration to Power Platform - How To and guidance please

This is a pretty complex ask so thx in advance

I also posted over in General Power Automate forum but have had no replies so posting here as well

Bottom line

I want to supply data from the Text Extraction AI model in Power Apps to a model/job in Azure Machine Learning Studio that uses K means clustering and return back values from a K-means clustering model to a Power App to determine what column text should be grouped into within a multi column text extraction from a page of text (image)

Ask

I can't find any examples/documentation how to do this. I am guessing I should use Power Automate and pass the data (left, width) from the text extract AI model to some endpoint associated with the model/job in azure Machine Learning Studio but I can't find any docs or examples how to pass data into a model/job in Azure Machine Learning from Power Apps or Power Automate Flow and get the output of the Azure Machine learning job back into my flow or Power App. Can anyone point me to anything that will help me sort out how to do this?

Background

I am building a multicolumn Text Extraction app in Power Apps using the Text Recognition AI connector in Power Apps. I am able to extract the text along with the boundingBox created by the ‘Text recognition” model for each block of text on a page. It has four properties that include the following (and it also has the extracted text)

left
right
height
width

I want to use the values for left and width values I am getting from the Text Recognition model for each block of text into a pipeline in Azure machine learning that uses K-Means clustering to determine the proper columns in a multi-column document that I scan with my phone.

I can extract columns manually at present (I have it all working) by letting the user set the left and width values manually along with a delta value to handle any small variations in each column - It works well and stores the output along with the image into a SharePoint Doc Library... but I would REALLY like to use K-Means clustering to assign the columns rather than having the user set values for the Left and Width clusters ... hope that makes sense...

Quick Links