Entry detection and recognition for custom forms

A computerized method for providing signature data of an input document, the computerized method comprising:

• identifying, by the processor, a first set of signature fields of the input document using signature key-value pairs of text data of the input document;
• determining, by the processor, a first subset of signed signature fields and a first subset of unsigned signature fields from the first set of signature fields based on mapping to a set of predicted values;
• determining, by the processor, a second set of signature fields of the input document using a region prediction model applied to image data of the input document;
• obtaining, by the processor, region images associated with each signature field of the first subset of unsigned signature fields and with each signature field of the second set of signature fields;
• determining, by the processor, a second subset of signed signature fields and a second subset of unsigned signature fields of the obtained region images using a signature recognition model, the signature recognition model trained using a dataset of synthetic signature fields generated for use as training data for the signature recognition model, the signature recognition model trained, using training data, to recognize a signature in a signature field using machine learning techniques;
• and
• providing, by the processor, signature output data of the input document, wherein the signature output data includes at least one of the following:
• the first subset of signed signature fields, the second subset of signed signature fields, and the second subset of unsigned signature fields.

One or more computer storage media having computer-executable instructions for providing entry field data of an input document that, upon execution by a processor, cause the processor to at least:

• identify a first set of entry fields of the input document using entry key-value pairs of text data of the input document;
• determine a first subset of filled entry fields and a first subset of unfilled entry fields from the first set of entry fields based on mapping to a set of predicted values;
• determine a second set of entry fields of the input document using a region prediction model applied to the image data of the input document;
• obtain region images associated with each entry field of the first subset of unfilled entry fields and with each entry field of the second set of entry fields;
• determine a second subset of filled entry fields and a second subset of unfilled entry fields of the obtained region images using an entry recognition model, the entry recognition model trained using a dataset of synthetic entry fields generated for use as training data for the entry recognition model, the entry recognition model trained, using the training data, to recognize an entry in an entry field using machine learning techniques;
• and
• provide entry output data of the input document, wherein the entry output data includes at least one of the following:
• the first subset of filled entry fields, the second subset of filled entry fields, and the second subset of unfilled entry fields.