Methods and systems of field detection in a document
Systems and methods are disclosed to receive a training data set comprising a plurality of document images, wherein each document image of the plurality of document images is associated with respective metadata identifying a document field containing a variable text; generate, by processing the plurality of document images, a first heat map represented by a data structure comprising a plurality of heat map elements corresponding to a plurality of document image pixels, wherein each heat map element stores a counter of a number of document images in which the document field contains a document image pixel associated with the heat map element; receive an input document image; and identify, within the input document image, a candidate region comprising the document field, wherein the candidate region comprises a plurality of input document image pixels corresponding to heat map elements satisfying a threshold condition.
1 . A method comprising:
receiving a plurality of document images, wherein each document image of the plurality of document images comprises a corresponding document field containing a variable text;
generating, by processing the plurality of document images, a first heat map represented by a data structure comprising a plurality of heat map elements corresponding to a plurality of document image pixels, wherein each heat map element stores a respective value specifying whether the corresponding document field contains a document image pixel associated with the heat map element;
receiving an input document image; and
identifying, using the first heat map, a candidate region of the input document image, the candidate region comprising the document field, wherein the candidate region comprises a plurality of input document image pixels corresponding to heat map elements satisfying a threshold condition.
2 . The method of claim 1 , wherein the candidate region is identified using a plurality of heat maps, the plurality of heat maps comprising the first heat map and one or more additional heat maps, wherein the each of the plurality of heat maps identify a potential document field location corresponding to the document field.
3 . The method of claim 2 , wherein each of the plurality of heat maps identify the potential document field location relative to a respective reference element in each of the plurality of heat maps.
4 . The method of claim 3 , wherein the respective reference element comprises one or more of a predefined word, or a predefined graphical element.
5 . The method of claim 2 , further comprising:
extracting a content of each document image of the plurality of document images, wherein the content is included in the potential document field location; and
analyzing the content of each document image using Byte Pair Encoding (BPE) tokens.
6 . The method of claim 5 , wherein analyzing the content comprises:
representing the content of each document image using BPE tokens to derive tokenized content for each document image;
generating vector representation of the tokenized content for each document image;
calculating a distance between a pair of vector representations of the tokenized content from two document images of the plurality of document images; and
in response to determining that the distance is less than a predefined value, indicating that the potential document field location is likely to be correct.
7 . The method of claim 1 , wherein each document image of the plurality of document images is associated with respective metadata defining a marked up document field location corresponding to the document field for each document image of the plurality of document images.
8 . A system comprising:
a memory device storing instructions;
a processing device coupled to the memory device, the processing device to execute the instructions to:
receive a plurality of document images, wherein each document image of the plurality of document images comprises a corresponding document field containing a variable text;
generate, by processing the plurality of document images, a first heat map represented by a data structure comprising a plurality of heat map elements corresponding to a plurality of document image pixels, wherein each heat map element stores a respective value specifying whether the corresponding document field contains a document image pixel associated with the heat map element;
receive an input document image; and
identify, using the first heat map, a candidate region of the input document image, the candidate region comprising the document field, wherein the candidate region comprises a plurality of input document image pixels corresponding to heat map elements satisfying a threshold condition.
9 . The system of claim 8 , wherein the candidate region is identified using a plurality of heat maps, the plurality of heat maps comprising the first heat map and one or more additional heat maps, wherein the each of the plurality of heat maps identify a potential document field location corresponding to the document field.
10 . The system of claim 9 , wherein each of the plurality of heat maps identify the potential document field location relative to a respective reference element in each of the plurality of heat maps.
11 . The system of claim 10 , wherein the respective reference element comprises one or more of a predefined word, or a predefined graphical element.
12 . The system of claim 9 , wherein the processing device is further to:
extract a content of each document image of the plurality of document images, wherein the content is included in the potential document field location; and
analyze the content of each document image using Byte Pair Encoding (BPE) tokens.
13 . The system of claim 12 , wherein to analyze the content, the processing device is to:
represent the content of each document image using BPE tokens to derive tokenized content for each document image;
generate vector representation of the tokenized content for each document image;
calculate a distance between a pair of vector representations of the tokenized content from two document images of the plurality of document images; and
in response to a determination that the distance is less than a predefined value, indicate that the potential document field location is likely to be correct.
14 . The system of claim 8 , wherein each document image of the plurality of document images is associated with respective metadata defining a marked up document field location corresponding to the document field for each document image of the plurality of document images.
15 . A non-transitory computer-readable storage medium comprising instructions that, when executed by a processing device, cause the processing device to:
receive a plurality of document images, wherein each document image of the plurality of document images comprises a corresponding document field containing a variable text;
generate, by processing the plurality of document images, a first heat map represented by a data structure comprising a plurality of heat map elements corresponding to a plurality of document image pixels, wherein each heat map element stores a respective value specifying whether the corresponding document field contains a document image pixel associated with the heat map element;
receive an input document image; and
identify, using the first heat map, a candidate region of the input document image, the candidate region comprising the document field, wherein the candidate region comprises a plurality of input document image pixels corresponding to heat map elements satisfying a threshold condition.
16 . The non-transitory computer-readable storage medium of claim 15 , wherein the candidate region is identified using a plurality of heat maps, the plurality of heat maps comprising the first heat map and one or more additional heat maps, wherein the each of the plurality of heat maps identify a potential document field location corresponding to the document field.
17 . The non-transitory computer-readable storage medium of claim 16 , wherein each of the plurality of heat maps identify the potential document field location relative to a respective reference element in each of the plurality of heat maps.
18 . The non-transitory computer-readable storage medium of claim 17 , wherein the respective reference element comprises one or more of a predefined word, or a predefined graphical element.
19 . The non-transitory computer-readable storage medium of claim 16 , wherein the processing device is further to:
extract a content of each document image of the plurality of document images, wherein the content is included in the potential document field location; and
analyze the content of each document image using Byte Pair Encoding (BPE) tokens.
20 . The non-transitory computer-readable storage medium of claim 15 , wherein each document image of the plurality of document images is associated with respective metadata defining a marked up document field location corresponding to the document field for each document image of the plurality of document images.