Identifying location of shreds on an imaged form
Disclosed herein is a machine learning application for automatically reading filled-in forms. There are multiple steps involved in using a computer to accurately read a handwritten form. First, the system identifies the form. Second, the system identifies what parts of the form are important. Third, the important parts are extracted as image data (known as shreds). Finally, fourth, the system interprets the shreds. This application is focused on steps two and three of that overall process. The disclosed techniques relate to training a machine learning system on a given series of forms such that when provided future filled-in forms within that series, the system is able to extract the portions of the filled-in form that are important/relevant.
1. A method comprising:
receiving image data of a plurality of filled-in forms having a first form type;
generating reduced image data by subtracting a set of pixels from each of the plurality of filled-in forms, the set of pixels are an inverted average of pixels within the image data;
determining pixel values comprising a pixel value for each pixel location in the reduced image data, wherein each pixel value of the pixel values represents a magnitude of variance of a corresponding pixel location in the reduced image data;
generating, based on the plurality of pixel values, an image representation of the reduced image data; and
calculating, based on the pixel values in the image representation, a plurality of coordinate sets of input fields in the first form type.
2. The method of claim 1 , further comprising:
shifting the image data into greyscale.
3. The method of claim 1 , further comprising:
inverting pixel values of the image data.
4. The method of claim 1 , further comprising:
averaging pixels of the image data of the plurality of filled-in forms by corresponding pixel location; and
reconstructing the plurality of filled-in forms from said averaging.
5. The method of claim 4 , wherein said subtracting pixels is of said reconstructing of the given filled-in form from the given filled-in form.
6. The method of claim 1 , further comprising:
blurring the reduced image data.
7. The method of claim 1 , wherein calculating a plurality of coordinate sets of input fields comprises:
identifying connected components of pixel values in the image representation;
calculating, based on the connected components, the plurality of coordinate sets of input fields.
8. The method of claim 7 , wherein said identifying connected components further comprises:
filtering out connected components based on any of:
percentage coverage of a page; or
percentage coverage of a given axis on the page.
9. The method of claim 1 , further comprising:
extracting image shreds from a given form of the plurality of filled-in forms based on the plurality of coordinate sets of input fields.
10. A system comprising:
a processor; and
a memory having instructions that when executed are configured to cause the processor to perform:
receiving image data of a plurality of filled-in forms having a first form type;
generating reduced image data by subtracting a set of pixels from each of the plurality of filled-in forms, the set of pixels are an inverted average of pixels within the image data;
determining pixel values comprising a pixel value for each pixel location in the reduced image data, wherein each pixel value of the pixel values represents a magnitude of variance of a corresponding pixel location in the reduced image data;
generating, based on the plurality of pixel values, an image representation of the reduced image data; and
calculating, based on the pixel values in the image representation, a plurality of coordinate sets of input fields in the first form type.
11. The system of claim 10 , where the memory further has instructions that when executed cause the processor to perform:
shifting the image data into greyscale.
12. The system of claim 10 , where the memory further has instructions that when executed cause the processor to perform:
inverting pixel values of the image data.
13. The system of claim 10 , where the memory further has instructions that when executed cause the processor to perform:
averaging pixels of the image data of the plurality of filled-in forms by corresponding pixel location; and
reconstructing the plurality of filled-in forms from said averaging.
14. The system of claim 13 , wherein said subtracting pixels is of said reconstructing of the given filled-in form from the given filled-in form.
15. The system of claim 10 , where the memory further has instructions that when executed cause the processor to perform:
blurring the reduced image data.
16. The system of claim 10 , wherein calculating a plurality of coordinate sets of input fields comprises:
identifying connected components of pixel values in the image representation; and
calculating, based on connected components, the plurality of coordinate sets of input fields.
17. The system of claim 16 , wherein said identifying connected components further comprises:
filtering out connected components based on any of:
percentage coverage of a page; or
percentage coverage of a given axis on the page.
18. The system of claim 10 , where the memory further has instructions that when executed cause the processor to perform:
extracting image shreds from a given form of the plurality of filled-in forms based on the plurality of coordinate sets of input fields.