Abstract: A method for measuring the components of an input image on which an image classifier bases its decision regarding the assignment of this input image to one or multiple class(es) of a predefined classification. The method includes: providing binary masks, which indicate which pixels of the input image and/or of an intermediate product formed in the image classifier are considered relevant; assessing the binary masks using a quality function, which is a measure of the extent to which at least one classification score, supplied by the image classifier, with respect to at least one target class changes when the pixels of the input image or of the intermediate product which are relevant according to the binary mask are changed; and ascertaining the sought-after components of the input image relevant for the decision of the image classifier from the combination of the binary masks with respective assessments by the quality function.

Abstract: The present disclosure involves systems, software, and computer implemented methods for transaction auditing. One example method includes determining valid pixel-based pattern(s) that are included in valid reference images. Fraudulent pixel-based pattern(s) that are included in fraudulent reference images are determined. A request to classify an image is received. A determination is made as to whether pixel values in the image match a valid pixel-based pattern or a fraudulent pixel-based pattern. In response to determining that the pixel values match a valid pixel-based pattern, a likelihood of classifying the first image as a valid image is increased. In response to determining that the pixel values match a fraudulent pixel-based pattern, a likelihood that the image as a fraudulent image is increased. The image is classified in response to the request as either a valid image or a fraudulent image based on the likelihoods.

Abstract: An image processing apparatus includes a display device configured to display information, a reading device configured to read a document, and one or more controllers configured to function as a unit configured to input an image read by the reading device to a trained model trained based on an image that does not contain text and orientation information about the image that does not contain text, and a unit configured to display information about the image read by the reading device on the display device based on at least an output result from the trained model.

Abstract: A display apparatus includes circuitry to display, on a display, an image including a table, receive an operation of specifying a range to be edited in the image, acquire coordinates of lines of the table in the range, and change a color of pixels other than pixels corresponding to the lines in the range to a predetermined color.

Abstract: Example methods and systems are described to add a watermark for printing in a virtual desktop environment having an agent side and a client side. A watermark can be configured at the agent side for printing at the client side. At the agent side, a fallback font can be determined for text of the watermark, and coordinate space calculation can be performed, so that the watermark prints correctly at the client side.

Abstract: Medical image segmentation using interactive refinement, in which the trained deep models are then utilized for the processing of medical imaging are described. Operating a two-step deep learning training framework including receiving original input images at the deep learning training framework; generating an initial prediction image specifying image segmentation by base segmentation model; receiving user input guidance signals; routing each of (i) the original input images, (ii) the initial prediction image, and (iii) the user input guidance signals to an InterCNN; generating a refined prediction image specifying refined image segmentation by processing each of the (i) the original input images, (ii) the initial prediction image, and (iii) the user input guidance signals through the InterCNN to render the refined prediction image incorporating the user input guidance signals; and outputting a refined segmentation mask to the deep learning training framework as a guidance signal.

Abstract: An image processing device a camera and a method for of encoding images captured by a camera are disclosed. For each image of an image sequence captured by the camera, the image is pre-processed by filtering the image by applying a privacy filter, the privacy filter being configured to distort the image in such a way that privacy is achieved in the filtered image, and, for at least a subset of the filtered images, by colour revising the filtered image by changing colour of pixels of a plurality of scattered areas of the filtered image such that a respective colour of one or more pixels of each area of the plurality of scattered areas represents one or more original colours of one or more pixels before filtering at a location of that area in the filtered image. The pre-processed images are then encoded into an encoded video stream.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for machine learning. One of the methods includes receiving one or more page images from a document; for each page image: providing the page image to a computer vision neural network model, wherein the neural network model is trained for the particular page type and is configured to output predictions of coordinates for one or more regions within the image and corresponding labels for the respective regions; and generating an output data structure associating each labeled region with text content located within the labeled region.

Abstract: A text detection process may involve comparing high-contrast pixel densities of areas of images of a video to determine quantities of text-containing areas in the images. Based on a difference between quantities of text-containing areas of subsets of the images, an image of the video may be selected for modification.

Abstract: The systems and methods described herein relate to improvements to generative artificial intelligence systems through the use of generative artificial intelligence pipelines to supply external information to pre-trained large language models for use in answering queries. To improve the efficiency and accuracy of large language models in responding to user queries, according to various aspects described herein, such queries may be modified and augmented with additional relevant information and may be divided into multiple queries for parallel handling, the results of which may then be combined into a response. The additional relevant information may include portions of documents or other data sets to be used in generating the response. Additional aspects may further improve resilience and flexibility by managing the generation or implementation of such modified and augmented queries.

Abstract: Computer vision systems and methods for text classification are provided. The system detects a plurality of text regions in an image and generates a bounding box for each detected text region. The system utilizes a neural network to recognize text present within each bounding box and classifies the recognized text, based on at least one extracted feature of each bounding box and the recognized text present within each bounding box, according to a plurality of predefined tags. The system can associate a key with a value and return a key-value pair for each predefined tag.

Abstract: An illustrative image processing system extracts a first color-field image from an original color image associated with a set of color-field components. The first color-field image is associated with a first subset of the set of color-field components. The image processing system also extracts a second color-field image from the original color image. The second color-field image is associated with a second subset of the set of color-field components that is different from the first subset. The image processing system detects a first set of features within the first color-field image and a second set of features within the second color-field image. At least one feature is detected within the first color-field image and included in the first set of features while not being detected within the second color-field image or included in the second set of features. Corresponding methods and systems are also disclosed.

Abstract: Systems, apparatuses, and methods for calculating multi-pass histograms for palette table derivation are disclosed. An encoder calculates a first histogram for a first portion of most significant bits (MSBs) of pixel component values of a block of an image or video frame. Then, the encoder selects a given number of the highest pixel count bins from the first histogram. The encoder then increases the granularity of these selected highest pixel count bins by evaluating one or more additional bits from the pixel component values. A second histogram is calculated for the concatenation of the original first portion MSBs from the highest pixel count bins and the one or more additional bits, and the highest pixel count bins are selected from the second histogram. A palette table is derived based on these highest pixel count bins selected from the second histogram, and the block is encoded using the palette table.

Abstract: A system for managing documents, comprising: interfaces to a user interface, proving an application programming interface, a database of document images, a remote server, configured to communicate a text representation of the document from the optical character recognition engine to the report server, and to receive from the remote server a classification of the document; and logic configured to receive commands from the user interface, and to apply the classifications received from the remote server to the document images through the interface to the database. A corresponding method is also provided.

Abstract: A device may receive image data representing a document, the document including: text, and edges. Based on the edges, the device may identify, a segment of interest within the image data and crop the segment of interest to obtain a portion of the image data. In addition, the device may perform optical character recognition on the portion of the image data, the optical character recognition producing recognized text. The device may obtain, based on the recognized text, validation data that includes verification text, and determine whether the recognized text is verified based on the verification text. Based on a result of the determination, the device may perform an action.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for neural network-based optical character recognition. An embodiment of the system may generate a set of bounding boxes based on reshaped image portions that correspond to image data of a source image. The system may merge any intersecting bounding boxes into a merged bounding box to generate a set of merged bounding boxes indicative of image data portions that likely portray one or more words. Each merged bounding box may be fed by the system into a neural network to identify one or more words of the source image represented in the respective merged bounding box. The one or more identified words may be displayed by the system according to a standardized font and a confidence score.

Abstract: An image processing system accesses an image of a completed form document. The image of the form document includes one or more features, such as form text, at particular locations within the image. The image processing system accesses a template of the form document and computes a rotation and zoom of the image of the form document relative to the template of the form document based on the locations of the features within the image of the form document relative to the locations of the corresponding features within the template of the form document. The image processing system performs a rotation operation and a zoom operation on the image of the form document, and extracts data entered into fields of the modified image of the form document. The extracted data can be then accessed or stored for subsequent use.

Abstract: An information processing apparatus (2000) displays, on a display apparatus (60), a first display (30) that represents a partial image (14) where an object contained therein was not recognized as a product. The information processing apparatus (2000) receives input for selecting one or more first displays (30). The information processing apparatus (2000), upon receiving a predetermined input from a user, ends receiving selection of a first display (30). The information processing apparatus (2000) stores, in a storage apparatus (120), product identification information input to a product information input area (54) and feature information based on a partial image(s) (14) corresponding to a selected first display(s) in association with each other.

Abstract: A first binary image is generated by binarizing an input image based on a threshold, a second binary image is generated by changing a pixel that has predetermined high luminance in the input image into a black pixel, and whether a black pixel cluster in the second binary image is made to be an extraction target is determined based on a position of a character image identified based on a black pixel cluster in the first binary image, and a position of the black pixel cluster in the second binary image.

Abstract: A system may include a computer readable medium and a processor communicatively coupled to the computer readable medium. The processor may be configured to: obtain a graphical image file, the graphical image file including an image, wherein the image includes at least one sequence of repeating pattern elements, each of the at least one sequence including the repeating pattern elements that are repeated along a linear direction; and convert the graphical image file to at least one file including hardware directives that when executed cause a recreation of the image of the graphical image file to be drawn, wherein a file size of the at least one file is smaller than the graphical image file.

Abstract: Computer-implemented methods and systems for colour-based image tagging and colour-based searching. The method may include identifying, using image analysis, one or more dominant colours of a product based on an image of the product and receiving selection of at least one of the one or more dominant colours. In response to receiving the selection of the at least one of the one or more dominant colours, a search for products matching the at least one of the one or more dominant colours may be initiated to obtain one or more results of the searching, the one or more results including at least one product matching the at least one of the one or more dominant colours.

Abstract: Disclosed herein is a document image analysis apparatus including: a document image acquisition unit configured to acquire a document image; a region detection unit configured to detect a plurality of regions from the document image acquired by the document image acquisition unit; a clustering unit configured to cluster the plurality of regions detected by the region detection unit to integrate into a cluster; and a reading order assignment unit configured to assign a reading order to a plurality of regions belonging to the cluster within the cluster integrated by the clustering unit.

Abstract: A method and apparatus for recognizing an imaged information-bearing medium, a computer-readable storage device and a computer device are provided. The method comprising: acquiring a first image of the imaged information-bearing medium; performing text recognition on the first image to acquire a text content of the imaged information-bearing medium; classifying the imaged information-bearing medium to acquire a type of the imaged information-bearing medium; and archiving the text content according to the type.

Abstract: One example method involves operations for a processing device that include receiving, by a machine learning model trained to generate a search result, a search query for a text input. The machine learning model is trained by receiving pre-training data that includes multiple documents. Pre-training the machine learning model by generating, using an encoder, feature embeddings for each of the documents included in the pre-training data. The feature embeddings are generated by applying a masking function to visual and textual features in the documents. Training the machine learning model also includes generating, using the feature embeddings, output features for the documents by concatenating the feature embeddings and applying a non-linear mapping to the feature embeddings. Training the machine learning model further includes applying a linear classifier to the output features. Additionally, operations include generating, for display, a search result using the machine learning model based on the input.

Abstract: This disclosure covers methods, non-transitory computer readable media, and systems analyze a digital design document having an initial layout of digital objects and automatically generate candidate layouts by concurrently performing operations on the digital objects within the initial layout. By iteratively performing concurrent operations, in some implementations, the methods, non-transitory computer readable media, and systems produce multiple candidate layouts that the systems evaluate by generating design scores. Based on a comparison of such design scores, the methods, non-transitory computer readable media, and systems generate one or more modified layouts (from among the candidate layouts) for presentation to a user.

Abstract: Systems and methods are provided for capturing by a camera of a user device, a first image depicting a first environment of the user device; overlaying a first virtual object on a portion of the first image depicting the first environment; modifying a surface of the first virtual object using content captured by the user device; storing a second virtual object comprising the first virtual object with the modified surface; and generating for display the second virtual object on a portion of a second image depicting a second environment.

Abstract: An automated communication design analysis and construction system that includes one or more intelligent communication design servers, comprising: a normalization module that converts communication content files for different recipients to normalized intermediate format files; an objects identification and quantification module that identifies text objects and image objects in the normalized intermediate format files; a cross-recipient group analysis module configured to identify static global objects that are invariant between recipients, data variables, and variable global objects that vary between recipients in the normalized intermediate format files; and an intelligent communication content learning and constructing engine that can construct standard communication design files based on the static global objects, the data variables, and the variable global objects. A data storage stores the communication content files and the standard communication design files.

Abstract: This disclosure relates generally to method and system for determining label value for labels in unstructured documents. Typical systems have challenge in understanding variations in layout of unstructured documents and extract information therefrom. The disclosed method and system facilitate systematically identifying sections and bounding boxes in the page images, taking image portion of the bounding boxes and extracting labels and label values therefrom. In case the label values are not present in the same bounding box having the label, the neighboring labels are examined for the matching label values. The system also obtains label-label value pairs from the document by utilizing a trained deep learning model, and compares the output with the label-label value pairs extracted earlier. An aggregated confidence score is assigned to the text in the bounding box.

Abstract: Image encoded documents are identified by recognizing known objects in each document with an object recognizer. The objects in each page are filtered to remove lower order objects. Known features in the objects are recognized by sequentially organizing each object in each filtered page into a one-dimensional array, where each object is positioned in a corresponding one-dimensional array as a function of location in the corresponding filtered page. The one-dimensional array is then compared to known arrays to classify the image document corresponding to the one-dimensional array.

Abstract: Digital content design system techniques are described using baseline units to control arrangement and sizing of digital content. In one example, a digital content design system receives a user input specifying a number of baselines to be included within an available display area of a page. Baselines are used to align digital content to control arrangement of the digital content within the page, e.g., text. From this, the digital content design system then calculates a baseline unit from a distance used to space adjacent baselines of the number of baselines from each other. This baseline unit is then leveraged by the system as a fundamental unit of measure to control arrangement and/or sizing of digital content in relation to each other.

Abstract: Examples for determining a confidence level associated with image segmentation are disclosed. A confidence level associated with a collective image segmentation result can be determined by generating multiple individual segmentation results each from the same image data. These examples can then aggregate the individual segmentation results to form the collective image segmentation result and measure the spread of each individual segmentation result from the collective image segmentation result. The measured spread of each individual segmentation result can then be used to determine the confidence level associated with the collective image segmentation result. This can allow a confidence level associated with the collective image segmentation result to be determined. This confidence level may be determined without needing a ground truth to compare to the collective image segmentation result.

Abstract: In implementations of systems for generating snap guides relative to glyphs of editable text rendered in a user interface using a font, a computing device implements a snap guide system to receive input data describing a position of a cursor relative to the glyphs of the editable text in the user interface. The glyphs of the editable text are enclosed within a bounding box having a height that is less than a height of an em-box of the font. The snap guide system generates a first group of snap guides for the glyphs of the editable text which includes a snap guide for each side of the bounding box and a snap guide for an x-height of the font. The snap guide system generates an indication of a particular snap guide of the first group of snap guides for display in the user interface based on the position of the cursor.

Abstract: A system may include a computer readable medium and a processor communicatively coupled to the computer readable medium. The processor may be configured to: obtain a graphical image file, the graphical image file including an image, wherein the image includes at least one sequence of repeating pattern elements, each of the at least one sequence including the repeating pattern elements that are repeated along a linear direction; and convert the graphical image file to at least one file including hardware directives that when executed cause a recreation of the image of the graphical image file to be drawn, wherein a file size of the at least one file is smaller than the graphical image file.

Abstract: Various embodiments disclosed herein are directed to methods of capturing Vehicle Identification Numbers (VIN) from images captured by a mobile device. Capturing VIN data can be useful in several applications, for example, insurance data capture applications. There are at least two types of images supported by this technology: (1) images of documents and (2) images of non-documents.

Abstract: A spatial indexing system receives a set of walkthrough videos of an environment taken over a period of time and receives an image search query that includes an image of an object. The spatial indexing system searches the set of walkthrough videos for instances of the object. The spatial indexing system presents search results in a user interface, displaying in a first portion a 2D map associated with one walkthrough video with marked locations of instances of the object and a second portion with a histogram of instances of the object over time in the set of walkthrough videos.

Abstract: In a collection technique, a user (such as a taxpayer) provides information (such as income-tax information) by submitting an image of a document, such as an income-tax summary or form. In particular, the user may provide a description of the document. In response, the user is prompted for the information associated with the field in the document. Then, the user provides the image of a region in the document that includes the field. Based on the image, the information is extracted, and the field in the form is populated using the extracted information. The prompting, receiving, extracting and populating operations may be repeated for one or more additional fields in the document.

Abstract: Disclosed are an apparatus and a method for detecting whether an anomalous sentence having a context different from that of other sentences exists in a document. The apparatus for detecting a contextually-anomalous sentence in a document according to the present invention includes: a sentence encoder for encoding individual sentences constituting document data by means of a predetermined rule (function) to generate encoding vectors; a context embedder neural network for converting the generated encoding vector into embedding vectors corresponding thereto; and a context anomaly detector neural network for detecting whether an anomalous sentence exists in the converted document data.

Abstract: In some embodiments, a method can include executing a first model to extract a first region of interest (ROI) image and a second ROI image from an image that shows an item and an indication of information associated to the item. The first ROI image can include a portion of the image showing the item and the second ROI image can include a portion of the image showing the indication of information. The method can further include executing a second model to identify the item from the first ROI image and generate a representation of the item. The method can further include executing a third model to read the indication of information associated to the item from the second ROI image and generate a representation of information.

Abstract: An automated method and system for validating (cross-validating) data fields in an electronic document, such as a document that has been passed through an optical character recognition (“OCR”) or Intelligent Document Recognition (“IDR”) system or software, to improve accuracy of the electronic document.

Abstract: An image processing engine and method of forming a hologram of a target image for projection using data streaming. An input or primary image is sub-sampled using a kernel and the secondary image output used to generate a hologram of the target image. A technique of kernel sub-sampling using a plurality of two or more data streams provides improvements in efficiency, including reduced data storage requirements and increased processing speed.

Abstract: Image data having text associated with a plurality of text-field types is received, the image data including target image data and context image data. The target image data including target text associated with a text-field type. The context image data providing a context for the target image data. A trained neural network that is constrained to a set of characters for the text-field type is applied to the image data. The trained neural network identifies the target text of the text-field type using a vector embedding that is based on learned patterns for recognizing the context provided by the context image data. One or more predicted characters are provided for the target text of the text-field type in response to identifying the target text using the trained neural network.

Abstract: Systems and methods for digitized document image data spillage recovery are provided. One or more memories may be coupled to one or more processors, the one or more memories including instructions operable to be executed by the one or more processors. The one or more processors may be configured to capture an image; process the image through at least a first pass to generate a first contour; remove a preprinted bounding region of the first contour to retain text; generate one or more pixel blobs by applying one or more filters to smudge the text; identify the one or more pixel blobs that straddle one or more boundaries of the first contour; resize the first contour to enclose spillage of the one or more pixel blobs; overlay the text from the image within the resized contour; and apply pixel masking to the resized contour.

Abstract: An online system extracts information from non-fixed form documents. The online system receives an image of a form document and obtains a set of phrases and locations of the set of phrases on the form image. For at least one field, the online system determines key scores for the set of phrases. The online system identifies a set of candidate values for the field from the set of identified phrases and identifies a set of neighbors for each candidate value from the set of identified phrases. The online system determines neighbor scores, where a neighbor score for a candidate value and a respective neighbor is determined based on the key score for the neighbor and a spatial relationship of the neighbor to the candidate value. The online system selects a candidate value and a respective neighbor based on the neighbor score as the value and key for the field.

Abstract: An image processing apparatus includes one or more processors; and a memory, the memory storing instructions, which when executed by the one or more processors, cause the one or more processors to generate vertical direction distribution data indicating a frequency distribution of distance values with respect to a vertical direction of a range image, from the range image having distance values according to distance of a road surface in a plurality of captured images captured by a plurality of imaging parts; set a search range corresponding to a predetermined reference point in the vertical direction distribution data and extract a plurality of pixels from the search range; and detect a road surface, based on the plurality of extracted pixels.

Abstract: The present application discloses a method and an apparatus for optical character recognition, an electronic device and a storage medium, and relates to the fields of artificial intelligence and deep learning. The method may include: determining, for a to-be-recognized image, a text bounding box of a text area therein, and extracting a text area image from the to-be-recognized image according to the text bounding box; determining a bounding box of text lines in the text area image, and extracting a text-line image from the text area image according to the bounding box; and performing text sequence recognition on the text-line image, and obtaining a recognition result. The application of the solution in the present application can improve a recognition speed and the like.

Abstract: Data processing systems (e.g. image processing systems) and methods are provided for processing a stream of data values (e.g. pixel values). In each of a plurality of iterations, a respective particular data value of the stream is processed by operating on a respective particular subset of data values of the stream. In each iteration: group indication data for at least one group is retrieved and used to define a set of groups into which data values within the particular subset can be grouped; each of the data values within the particular subset is grouped into one of the groups of the set of groups; the particular data value is processed using one or more of the data values of the particular subset in dependence on the classification of the data values into the groups; and group indication data is stored for a group, for use in a subsequent iteration.

Abstract: An information processing apparatus includes a memory and a processor configured to acquire an image of a digitized document and execute a first verification process by using image processing using artificial intelligence. The first verification process verifies whether a first requirement is satisfied. The first requirement is a specific requirement among multiple requirements that are required when the acquired image of the document is stored. The processor is also configured to execute a second verification process by using a determination process not using the artificial intelligence. The second verification process verifies whether a second requirement among the multiple requirements is satisfied. The second requirement is other than the first requirement.

Abstract: In a method for generating synthetic medical image data, first image data of an object under examination including a first value for a property is acquired, second image data of the object under examination including a second value for the property is acquired, the second value of the property of the second image data is matched to the first value to modify the second image data to generate synthetic image data, and the synthetic image data is provided (e.g. in electronic form as a data file). The first image data can be captured with a first magnetic resonance device at a first point in time, and the second image data can be captured with a second magnetic resonance device at a second point in time.

Abstract: A method is provided for encoding a digital video to provide for improved color mapping.

Abstract: Systems, methods, and products for auto tagging structured PDF documents that do not have accessibility tags. In one embodiment, structured PDF documents having accessibility tags are first parsed and analyzed to organize the visual components of the documents. The relationships of the identified objects to DOM elements (e.g., tags) are determined, and the objects and related DOM elements are stored in training files. The training files are used to train various classifiers. Untagged PDF documents are then parsed to identify included visual objects, and the classifiers are used to determine DOM elements that should be associated with visual objects identified in the untagged PDF documents. This information is used to construct a DOM structure corresponding to each untagged document. A new PDF is then generated corresponding to each untagged document using the generated DOM structure and visual object information.