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: A Fast Non-Maximum Suppression (NMS) Algorithm post-processing for object detection includes getting original data output from a deep learning model inference output, the original data including a plurality of bounding boxes, pre-emptively filtering out at least one bounding box of the plurality of bounding boxes from further consideration when applying the algorithm, the at least one bounding box filtered out according to a predetermined criteria, processing data, using sigmoid functions or exponential functions, from bounding boxes of the plurality of bounding boxes not filtered out to generate processed bounding boxes, calculating final scores of the processed bounding boxes, and choosing a processed bounding boxes utilizing the final scores.

Abstract: Systems and methods of generating synthetic training data for machine learning models. First, line items in source documents such as bills, invoices, and or receipts are identified and labeled. The identification and labeling generate labeled documents. Then, in the labeled documents, the line items are augmented by adding, deleting, and or swapping line items to generate synthetic training documents. An addition operation randomly selects one or more line items and adds the selected line item(s) to the same labeled document or another labeled document. A deletion operation randomly deletes one or more line items. A swapping operation randomly swaps line items in a single labeled document or across different labeled documents. These operations can generate synthetic labeled documents of any length, which form synthetic training data for training the machine learning models.

Abstract: A computing system engages in digital image processing of received video frames to generate sport data that indicates a score and/or a time associated with a sport event. The digital image processing includes: (i) identifying a first frame region of the video frames based on the first frame region depicting a scoreboard; (ii) executing a first procedure that analyzes the identified first frame region to detect, within the identified first frame region, second frame region(s) based on the second frame region(s) depicting text of the scoreboard; (iii) in response to detecting the second frame region(s), executing a second procedure to recognize the text in at least one of the second frame region(s); and (iv) based at least on the recognizing of the text, generating the sport data. In response to completing the digital image processing, the computing system then carries out an action based on the generated sport data.

Abstract: Provided herein are system, apparatus, device, method, and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for classifying a document using image analysis. In an embodiment, a server may train and implement a deep learning model to classify documents using image representations of the documents.

Abstract: Embodiments of the present disclosure provide systems, methods, and computer storage media for implementing a nuanced-color search and recommendation system. Employing various techniques described herein, an improved search engine for categorizable items is provided to generate accurate color-based search results and recommendations that are relevant to a nuanced color provided as a search parameter. In various implementations, digital assets that depict categorizable items can be analyzed to characterize the depicted items with, among other things, a particular shade of a determined dominant color. In this way, categorizable items can be searched based at least in part on a selected shade of a particular dominant color to generate a primary search result. In further embodiments, other categorizable items can be generated as recommendations based on characteristics of the primary search result.

Abstract: Method and system of automatically extracting information of a predefined type from a document is provided. The method includes identifying a location and classification of a segment of interest of a document that includes information associated with a predefined type. The method further includes identifying a location and classification of characters from the segment of interest based on characteristics associated with the predefined type. The method further includes extracting the identified characters from the segment of interested associated with the predefined type.

Abstract: Some embodiments are directed to, by a simple operation, appropriately displaying an image to which attention is paid. An information processing apparatus according to some embodiments is an information processing apparatus for performing display for measuring a width of a region in an image of an inspection target object, the information processing apparatus including an acquisition unit configured to acquire a bending position where a bend of a line approximating the region from the image satisfies a predetermined condition, and a display control unit configured to display, in an aligned manner in a predetermined direction, a plurality of partial images obtained by dividing the image based on the bending position.

Abstract: An approach to identifying text within an image may be presented. The approach can receive an image. The approach can classify an image on a pixel-by-pixel basis whether the pixel is text. The approach can generate bounding boxes around groups of pixels that are classified as text. The approach can mask sections of an image that where pixels are not classified as text. The approach may be used as a pre-processing technique for optical character recognition in documents, scanned images, or still images.

Abstract: Aspects of the disclosure relate to OCR. A computing platform may train, using historical images, a CNN and a RNN to perform OCR/identify characters in context. The computing platform may receive an image of a document, and may input the image into the CNN, which may cause the CNN to output OCR information for the image and a confidence score. Based on identifying that the confidence score exceeds a confidence threshold, the computing platform may store the OCR information to enable subsequent access of a digital version of the document. Based on identifying that the confidence score does not exceed the confidence threshold, the computing platform may: 1) input the OCR information into the first RNN, which may cause the first RNN to output contextual OCR information for the image, and 2) store the contextual OCR information to enable subsequent access of the digital version of the document.

Abstract: A method of blank detection involves receiving a document from a user, where the document includes derived text; applying a trained blank detection model to the document to make a first prediction, where the first prediction indicates whether at least one field in the document is blank; comparing the first prediction with a second prediction, where the second prediction is made by an extraction model; and extracting the at least one field using the extraction model.

Abstract: Selected artifacts, which may be based on distortions or selected attenuation features, may be reduced or removed from a reconstructed image. Various artifacts may occur due to the presence of a metal object in a field of view. The metal object may be identified and removed from a data that is used to generate a reconstruction.

Abstract: In an embodiment, operations include extracting first information about a first set of features of a first candidate, from a document or profile information of the first candidate. Second information about a second set of features, corresponding to the first set of features, is extracted from one or more databases. The second set of features is associated with a population of candidates with at least one demographic parameter same as that of the first candidate. A third set of features is determined based on difference of corresponding features from the first set of features and the second set of features. A pre-trained neural network model is applied on the third set of features to determine a set of weights associated with the third set of features. An empathy score of the first candidate is determined based on the set of weights. The empathy score of the first candidate is rendered.

Abstract: A device for automating the process of installing eyelash extensions onto the natural eyelashes of a subject. In some embodiments, the placing of extensions is carried out by a robotic mechanism utilizing computer vision. In some embodiments, the same device may be used for extending eyebrows rather than eyelashes, as eyebrows have rather similar properties to eyelashes.

Abstract: Aspects of the disclosure provide for mechanisms for identification of blocks of associated words in documents using neural networks. A method of the disclosure includes obtaining a plurality of words of a document, the document having a first block of associated words, determining a plurality of vectors representative of the plurality of words, processing the plurality of vectors using a first neural network to obtain a plurality of recalculated vectors having values based on the plurality of vectors, determining a plurality of association values corresponding to a connections between at least two words of the document, and identifying, using the plurality of recalculated vectors and the plurality of association values, the first block of associated symbol sequences.

Abstract: A method for controlling a robotic device is presented. The method includes capturing an image corresponding to a current view of the robotic device. The method also includes identifying a keyframe image comprising a first set of pixels matching a second set of pixels of the image. The method further includes performing, by the robotic device, a task corresponding to the keyframe image.

Abstract: Methods, systems, and computer program products for vision-based cell structure recognition using hierarchical neural networks and cell boundaries to structure clustering are provided herein. A computer-implemented method includes detecting a style of the given table using at least one style classification model; selecting, based at least in part on the detected style, a cell detection model appropriate for the detected style; detecting cells within the given table using the selected cell detection model; and outputting, to at least one user, information pertaining to the detected cells comprising image coordinates of one or more bounding boxes associated with the detected cells.

Abstract: An optical device for capturing an image of a scene, the optical device comprising: a plurality of image sensors each operable to capture a respective initial image of the scene; a lens arrangement operable to receive light from the scene and to form each initial image on each respective image sensor, each image sensor being located at a different respective distance from the lens arrangement; and an image processor operable to generate the captured image of the scene on the basis of image data from one or more of the captured initial images.

Abstract: An image processing method and apparatus. The method includes: obtaining a source image; determining spot superposition positions according to pixel values of the source image, where values of pixels of the source image that are located in the spot superposition positions are greater than a preset first threshold; and blurring the source image, and performing, in the spot superposition positions of a source image, image fusion on the source image and spot images to obtain a processed image, where the spot superposition positions and the spot images fused with in the spot superposition positions are in a one-to-one correspondence.

Abstract: A system and method for detecting construction site defects and hazards using artificial intelligence (AI) is provided. The system includes a movable base unit, a coordinate measurement scanner, a vision based sensor, and one or more processors. The one or more processors perform operations that include generating a two-dimensional (2D) map of the environment based at least in part on output from the coordinate measurement scanner, applying image recognition to the video stream data to identify and label a defect or hazard in the video data stream, correlating a location of the defect or hazard in the video stream data with the location in the 2D map, and recording the location of the defect or hazard in the 2D map.