FAST IDENTIFICATION OF IMAGES IN DOCUMENTS
Methods and systems for detecting images in documents are described. A method implemented by an electronic device having one or more processors for determining whether a document is an image includes partitioning a document into a plurality of cells. The method includes scaling each of the cells to a standardized number of pixels to provide a corresponding snippet for each of the cells, classifying the snippets, using a neural network, to determine a set of cells classified as text, and determining a volume of text for the document based on a sum of an amount of text in each cell of the set of cells. The method further includes in response to a determination that the volume of text for the document is below a predetermined threshold, determining that the document is an image.
1 . A method implemented by an electronic device having one or more processors for determining whether a document is an image, the method comprising:
partitioning a document into a plurality of cells;
scaling each of the cells to a standardized number of pixels to provide a corresponding snippet for each of the cells;
classifying the snippets, using a neural network, to determine a set of cells classified as text;
determining a volume of text for the document based on a sum of an amount of text in each cell of the set of cells; and
in response to a determination that the volume of text for the document is below a predetermined threshold, determining that the document is an image.
2 . The method of claim 1 , further comprising:
in response to a determination that the volume of text for the document is below the predetermined threshold:
classifying the snippets, using the neural network, to determine another set of cells classified as non-text;
in accordance with a determination that the other set of cells meet partitioning criteria, partitioning the other set of cells to form partitioned cells;
scaling each of the partitioned cells to a standardized number of pixels to provide a respective snippet for each of the partitioned cells;
classifying the respective snippets, using the neural network, to determine a set of partitioned cells classified as text;
determining an updated volume of text for the document based on a sum of an amount of text in each cell of the set of partitioned cells and the volume of text for the document; and
in response to a determination that the updated volume of text for the document is below the predetermined threshold, determining that the document is an image.
3 . The method of claim 2 , further comprising:
determining, based on the set of partitioned cells classified as text, that a portion of the document includes text.
4 . The method of claim 3 , further comprising:
in accordance with a determination that the document includes a predetermined amount of text, performing text-related processing on the document.
5 . The method of claim 2 , further comprising:
in response to a determination that the updated volume of text for the document is below the predetermined threshold:
classifying the snippets, using the neural network, to determine another set of partitioned cells classified as non-text;
in accordance with a determination that the other set of partitioned cells does not meet partitioning criteria, determining whether the set of cells and the set of partitioned cells have a satisfactory geometry; and
in response to a determination that the set of cells and the set of partitioned cells do not have a satisfactory geometry, determining that the document is an image.
6 . The method of claim 5 , further comprising:
in response to a determination that the set of cells and the set of partitioned cells have a satisfactory geometry, determining that the document is a text page.
7 . The method of claim 2 , wherein the respective snippets are classified in random order.
8 . The method of claim 2 , wherein the respective snippets are classified in an order that prioritizes respective snippets adjacent to snippets previously classified as text.
9 . The method of claim 1 , wherein one or more cells of the set of cells are aligned to form at least one text line and wherein the at least one text line is one of: horizontal or vertical.
10 . The method of claim 2 , wherein one or more cells of the other set of cells are classified as one of an image or unknown.
11 . The method of claim 2 , wherein partitioning the other set of cells to form the partitioned set of cells includes partitioning respective cells of the other set of cells into four cells.
12 . The method of claim 1 , wherein the document is captured using a smartphone.
13 . The method of claim 1 , wherein the neural network is trained using a plurality of image documents and a plurality of text pages having various formats, layouts, text sizes, ranges of word, line and paragraph spacing.
14 . A non-transitory computer readable medium storing one or more programs, the one or more programs comprising instructions, which when executed by a device with a camera, cause the device to:
partition a document into a plurality of cells;
scale each of the cells to a standardized number of pixels to provide a corresponding snippet for each of the cells;
classify the snippets, using a neural network, to determine a set of cells classified as text;
determine a volume of text for the document based on a sum of an amount of text in each cell of the set of cells; and
in response to a determination that the volume of text for the document is below a predetermined threshold, determine that the document is an image.
15 . The non-transitory computer readable medium of claim 14 , wherein the one or more programs further comprising instructions, which when executed by the device, cause the device to:
in response to a determination that the volume of text for the document is below a predetermined threshold:
classify the snippets, using the neural network, to determine another set of cells classified as non-text;
in accordance with a determination that the other set of cells meet partitioning criteria, partition the other set of cells to form partitioned cells;
scale each of the partitioned cells to a standardized number of pixels to provide a respective snippet for each of the partitioned cells;
classify the respective snippets, using the neural network, to determine a set of partitioned cells classified as text;
determine an updated volume of text for the document based on a sum of an amount of text in each cell of the set of partitioned cells and the volume of text for the document; and
in response to a determination that the updated volume of text for the document is below the predetermined threshold, determine that the document is an image.
16 . The non-transitory computer readable medium of claim 15 , wherein the one or more programs further comprising instructions, which when executed by the device, cause the device to:
determine, based on the set of partitioned cells classified as text, that a portion of the document includes text.
17 . The non-transitory computer readable medium of claim 16 , wherein the one or more programs further comprising instructions, which when executed by the device, cause the device to:
in accordance with a determination that the document includes a predetermined amount of text, perform text-related processing on the document.
18 . A device with a camera, the device comprising:
one or more processors; and
memory storing one or more instructions that, when executed by the one or more processors, cause the device to perform operations including:
partitioning a document into a plurality of cells;
scaling each of the cells to a standardized number of pixels to provide a corresponding snippet for each of the cells;
classifying the snippets, using a neural network, to determine a set of cells classified as text;
determining a volume of text for the document based on a sum of an amount of text in each cell of the set of cells; and
in response to a determination that the volume of text for the document is below a predetermined threshold, determining that the document is an image.
19 . The device of claim 18 , wherein one or more instructions, when executed by the one or more processors, cause the device to further perform operations including:
in response to a determination that the volume of text for the document is below a predetermined threshold:
classifying the snippets, using the neural network, to determine another set of cells classified as non-text;
in accordance with a determination that the other set of cells meet partitioning criteria, partitioning the other set of cells to form partitioned cells;
scaling each of the partitioned cells to a standardized number of pixels to provide a respective snippet for each of the partitioned cells;
classifying the respective snippets, using the neural network, to determine a set of partitioned cells classified as text;
determining an updated volume of text for the document based on a sum of an amount of text in each cell of the set of partitioned cells and the volume of text for the document; and
in response to a determination that the updated volume of text for the document is below the predetermined threshold, determining that the document is an image.
20 . The device of claim 19 , wherein one or more instructions, when executed by the one or more processors, cause the device to further perform operations including:
determining, based on the set of partitioned cells classified as text, that a portion of the document includes text.