Computer Vision for Prevention Point Forms

Computer Vision for Prevention Point Forms

Prevention Point collects data via handwritten forms. This project seeks to use computer vision to eliminate the need to do data entry by hand.

This project is to design a piece of software to record information from scanned forms for Prevention Point. This project utilizes openCV and the Microsoft Azure API. Please contact the project lead on slack on how to get access the API key.

Main Project Parts

1. Form Intake Interface

There are several different types of forms that need to be scanned. The form intake interface should allow the user to select which type of form is being scanned.

2. Rotate and Rescale Forms

In order for the bounding box section of the project to work. The images for each form need to be in the same orientation and scale. The current thinking is to use the prevention point logo on each page to orient and scale.

3. Create Proper Bounding Boxes to Collect Data from the Forms

For each of the forms bounding boxes need to be created around text fields and marks. The collected data should be deposited in a csv that corresponds to the variables in the variable dictionary for that form.

4. Flagging System

The client (Prevention Point) has flagged "Name" and "UniqueID" as the fields that need the most accurate information. The Microsoft Azure Read API returns a certainty number. If the certainty number is below a certain threshold. The software should display the form and a field that allows for manual entry and correction.