I am looking for a tool that can accept a PDF with fillable form fields and then detects the labels and section according to the layout of the document.

For example, in the document below, the form field highlighted in red should have label: “Policyholder(s) details -> Full name”. And the blue form field should have label: “Person driving or in charge of the vehicle (to be completed, even if parked) -> Business”.

Otherwise, if no such tool exists, how can I get started training a model that can achieve this? Is there any base model that I can train on? I’m still quite new to machine learning.