Abstract: An information processing apparatus includes a processor configured to receive an input image including images of plural documents, execute detection of one or more items determined in advance as an item included in the document from the input image, and execute output processing of extracting and outputting the image of each document from the input image based on the detected one or more items.

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: Embodiments relate to systems and methods to optimize quantization of tensors of an AI model. According to one embodiment, a system receives an AI model having one or more layers. The system receives a number of input data for offline inferencing and applies offline inferencing to the AI model based on the input data to generate offline data distributions for the AI model. The system quantizes one or more tensors of the AI model based on the offline data distributions to generate a low-bit representation AI model, where each layer of the AI model includes the one or more tensors, where the one or more tensors include the one or more tensors. In one embodiment, the system applies online inferencing using the low-bit representation AI model to generate online data distributions for a feature map, and quantizes a feature map tensor based on the online data distributions.

Abstract: A method of extracting information from a flowchart image comprising a plurality of closed-shaped data nodes having text enclosed within, connecting lines connecting the plurality of closed-shaped data nodes and free text adjacent to the connecting lines includes receiving the flowchart image, detecting the closed-shaped data nodes, localizing the text enclosed within the closed-shaped data nodes, and masking the localized text.to generate an annotated image. Lines in the annotated image are the detected to reconstruct them as closed-shaped data nodes and connecting lines. A tree frame with the plurality of closed-shaped data nodes and the connecting lines is extracted. The free text is then localized. Chunks of the free text oriented and positioned proximally together are assembled into text blocks using an orientation-based two-dimensional clustering.

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: A method, an electronic device and computer readable medium for entity-relationship embeddings using automatically generated entity graphs instead of a traditional knowledge graph are provided. The method includes receiving, by a processor, an input text. The method also includes identifying a primary entity, a secondary entity and a context from the input text, wherein the context comprises a relationship between the primary entity and the secondary entity. The method additionally includes generating, by the processor, an entity context graph based on the primary entity, the secondary entity, and the context by: extracting, from the context, one or more text segments comprising a plurality of words describing one or more additional relationships between the primary entity and the secondary entity, and generating a plurality of context triples from the one or more text segments, each of the plurality of context triples defining a respective relationship between primary entity and the secondary entity.

Abstract: The invention provides a method and a device for identifying the number of bills and multiple bill areas in an image. The method comprises the following steps: acquiring the image containing a plurality of bills arranged in sequence; processing the image to obtain a plurality of boundary lines of each bill in the image; wherein the boundary lines comprise a first type of boundary lines which are substantially perpendicular to the bill arrangement direction; generating a long line segment which is substantially parallel to the arrangement direction of the bills and passes through the area where all the bills are located, wherein the long line segment has an intersection point with each first type of boundary line; and determining the number of bills in the image according to the lengths of the sub-line segments between the adjacent intersection points.

Abstract: Data is received that is derived from each of a plurality of inspection camera modules forming part of a quality assurance inspection system. The data includes a feed of images of a plurality of objects passing in front of the respective inspection camera module. Thereafter, the received data is separately analyzed by each inspection camera module using at least one image analysis inspection tool. The results of the analyzing can be correlated for each inspection camera module on an object-by-object basis. The correlating can use timestamps for the images and/or detected unique identifiers within the images and can be performed by a cloud-based server and/or a local edge computer. Access to the correlated results can be provided to a consuming application or process.

Abstract: A system includes a computing platform having a hardware processor and a memory storing a software code and a neural network (NN) having multiple layers including a last activation layer and a loss layer. The hardware processor executes the software code to identify different combinations of layers for testing the NN, each combination including candidate function(s) for the last activation layer and candidate function(s) for the loss layer. For each different combination, the software code configures the NN based on the combination, inputs, into the configured NN, a training dataset including multiple data objects, receives, from the configured NN, a classification of the data objects, and generates a performance assessment for the combination based on the classification. The software code determines a preferred combination of layers for the NN including selected candidate functions for the last activation layer and the loss layer, based on a comparison of the performance assessments.

Abstract: A system reinforcement learning method includes: processing an input image based on a first network of a system to obtain a first result; inputting the first result to a second network of the system to obtain a second result; and obtaining a reinforcement operation based on the second result by means of a reinforcement network, and adjusting the first result based on the reinforcement operation to obtain a target result. According to the embodiments of the present disclosure, information is fed back from downstream to upstream by means of the reinforcement network, and an output result of the system is optimized.

Abstract: A workflow support apparatus includes a classification section that classifies a document included in an original document from image data acquired by reading the original document, and a workflow searching section that searches for a workflow to which the document is to be attached, from the document classified by the classification section.

Abstract: An image processing component is trained to process 2D images of human body parts, in order to extract depth information about the human body parts captured therein. Image processing parameters are learned during the training from a training set of captured 3D training images, each 3D training image of a human body part and captured using 3D image capture equipment and comprising 2D image data and corresponding depth data, by: processing the 2D image data of each 3D training image according to the image processing parameters, so as to compute an image processing output for comparison with the corresponding depth data of that 3D image, and adapting the image processing parameters in order to match the image processing outputs to the corresponding depth data, thereby training the image processing component to extract depth information from 2D images of human body parts.

Abstract: The techniques described herein relate to methods, apparatus, and computer readable media configured to test a pose of a three-dimensional model. A three-dimensional model is stored, the three dimensional model comprising a set of probes. Three-dimensional data of an object is received, the three-dimensional data comprising a set of data entries. The three-dimensional data is converted into a set of fields, comprising generating a first field comprising a first set of values, where each value of the first set of values is indicative of a first characteristic of an associated one or more data entries from the set of data entries, and generating a second field comprising a second set of values, where each second value of the second set of values is indicative of a second characteristic of an associated one or more data entries from the set of data entries, wherein the second characteristic is different than the first characteristic.

Abstract: Embodiments of the present invention provide a system that can be used to classify a feedback image in a user review into a semantically meaningful class. During operation, the system analyzes the captions of feedback images in a set of user reviews and determines a set of training labels from the captions. The system then trains an image classifier with the set of training labels and the feedback images. Subsequently, the system generates a signature for a respective feedback image in a new set of user reviews using the image classifier. The signature indicates a likelihood of the image matching a respective label in the set of training labels. Based on the signature, the system can allocate the image to an image cluster.

Abstract: The need for extracting information trapped in unstructured document images is becoming more acute. A major hurdle to this objective is that these images often contain information in the form of tables and extracting data from tabular sub-images presents a unique set of challenges. Embodiments of the present disclosure provide systems and methods that implement a deep learning network for both table detection and structure recognition, wherein interdependence between table detection and table structure recognition are exploited to segment out the table and column regions. This is followed by semantic rule-based row extraction from the identified tabular sub-regions.

Abstract: A method for employing a differentiable ranking based graph sparsification (DRGS) network to use supervision signals from downstream tasks to guide graph sparsification is presented. The method includes, in a training phase, generating node representations by neighborhood aggregation operators, generating sparsified subgraphs by top-k neighbor sampling from a learned neighborhood ranking distribution, feeding the sparsified subgraphs to a task, generating a prediction, and collecting a prediction error to update parameters in the generating and feeding steps to minimize an error, and, in a testing phase, generating node representations by neighborhood aggregation operators related to testing data, generating sparsified subgraphs by top-k neighbor sampling from a learned neighborhood ranking distribution related to the testing data, feeding the sparsified subgraphs related to the testing data to a task, and outputting prediction results to a visualization device.

Abstract: Embodiments of the present disclosure relate to a method and apparatus for recognizing a table, a device, and a medium. An embodiment of the method can include: detecting a table on a target picture, to obtain a candidate table recognition result; extracting a merging feature of the candidate table recognition result, and determining a to-be-merged row in the candidate table recognition result based on the merging feature; extracting a direction feature of the to-be-merged row, and determining a merging direction of the to-be-merged row based on the direction feature; and adjusting the candidate table recognition result based on the to-be-merged row and the merging direction of the to-be-merged row, to obtain a target table recognition result.

Abstract: Detection of typed and/or pasted text, caret tracking, and active element detection for a computing system are disclosed. The location on the screen associated with a computing system where the user has been typing or pasting text, potentially including hot keys or other keys that do not cause visible characters to appear, can be identified and the physical position on the screen where typing or pasting occurred can be provided based on the current resolution of where one or more characters appeared, where the cursor was blinking, or both. This can be done by identifying locations on the screen where changes occurred and performing text recognition and/or caret detection on these locations. The physical position of the typing or pasting activity allows determination of an active or focused element in an application displayed on the screen.

Abstract: Detection of typed and/or pasted text, caret tracking, and active element detection for a computing system are disclosed. The location on the screen associated with a computing system where the user has been typing or pasting text, potentially including hot keys or other keys that do not cause visible characters to appear, can be identified and the physical position on the screen where typing or pasting occurred can be provided based on the current resolution of where one or more characters appeared, where the cursor was blinking, or both. This can be done by identifying locations on the screen where changes occurred and performing text recognition and/or caret detection on these locations. The physical position of the typing or pasting activity allows determination of an active or focused element in an application displayed on the screen.

Abstract: A system for classification of scholastic works includes a computing device configured to receive a first scholastic work, identify an author and a category of the first scholastic work, determine at least a work theme by receiving theme training data, the theme training data including a plurality of entries, each entry including a training textual element and a correlated theme, training a theme classifier as a function of the training data, and determining the at least a work theme as a function of the plurality of textual elements and the theme classifier, calculate a reliability quantifier as a function of the at least a theme, the author, and the category, select the scholastic work as a function of the reliability quantifier, derive, from the scholastic work, at least a correlation between a dietary practice and alleviation of a disease state, and store the at least a correlation in an expert database.