OpenCV/Tensorflow - Train a machine to identify patterns in images

** I WILL DECLINE AUTO-BOT OFFERINGS FROM PEOPLE THAT DONT READ THE PROJECT DESCRIPTION **

Opportunity for you to help us train a machine learning model that can classify visual patterns in newspaper articles.

We need someone to create a command-line tool (using Python, Java, or other suitable programming language) that can detect the coordinates and regions of each article within a given newspaper page. This tool will be based on a large dataset of PDF/JPG newspaper pages that we will provide.

To give you an idea of what we're looking for, we've attached some sample newspaper pages that illustrate the kind of visual patterns we're interested in. Your tool should be able to analyze these pages and generate an output in JSON format (or XML) that specifies the coordinates and dimensions of each article region.

If you're up for the challenge, we'd love to hear from you. This is a great opportunity to demonstrate your skills and make a real impact in the field of machine learning.

Example: Input= the samples attached
Output:
[
{
"x":491,
"y": 230,
"width":330,
"height":550
},
{
"x":420,
"y": 330,
"width":530,
"height":150
}
...
]

If you are interested, don't send me just a generic message, I won't anwer - Explain how you would solve the problem.