Abstract: An artificial neural network-based method for detecting a surface pattern of an object includes receiving a plurality of object images, dividing each object image into a plurality of image areas, designating at least one region of interest from the plurality of image areas of each of the object images, and performing deep learning with the at least one region of interest to build a predictive model for identifying a surface pattern of the object.

Abstract: System, methods, and other embodiments described herein relate to self-supervised training for monocular depth estimation. In one embodiment, a method includes filtering disfavored images from first training data to produce second training data that is a subsampled version of the first training data. The disfavored images correspond with anomaly maps within a set of depth maps. The first depth model is trained according to the first training data and generates the depth maps from the first training data after initially being trained with the first training data. The method includes training a second depth model according to a self-supervised training process using the second training data. The method includes providing the second depth model to infer distances from monocular images.

Abstract: An apparatus and method are provided and include an interface unit configured to receive an electrical signal based on light received from a camera; and at least one processor configured to obtain depth information of a scene based on the electrical signal received from the interface unit, wherein the at least one processor is configured to identify a recognition pattern corresponding to the light received by the camera based on the electrical signal, to obtain a total distance value between properties of a partial source pattern corresponding to a target fragment in the source pattern and properties of a partial recognition pattern corresponding to the target fragment in the recognition pattern, and to estimate depth information of the target fragment based on the obtained total distance value.

Abstract: A calibration system includes a processing circuit configured to receive data corresponding to a first image where the data includes information corresponding to a 2D BEV of a scene, and to receive data corresponding to a second image where the data includes information corresponding to a 2D image of the scene. The processing circuit may determine a salient feature in the first image or the second image, and project, based on a first calibration setting, the salient feature from the first image to the second image or vice versa. The processing circuit may select a region corresponding to the projected salient feature in the first image or the second image, determine an identity of the salient feature, and determine a second calibration setting based on the identity of the salient feature and the selected region corresponding to the salient feature in the 2D BEV and the 2D image.

Abstract: Devices, methods, and computer-readable media describing an adaptive approach for image selection, fusion, and noise reduction, e.g., to generate low noise and high dynamic range (HDR) images with improved motion freezing in a variety of capturing conditions. An incoming image stream may be obtained from an image capture device, wherein the image stream comprises a variety of differently-exposed captures, e.g., EV0 images, EV? images, EV+ images. When a capture request is received, a set of rules may be used to evaluate one or more capture conditions associated with the images from the incoming image stream and determine which two or more images to select for a fusion operation. The fusion operation may be designed to adaptively fuse the selected images, e.g., in a fashion that is determined to be optimal from a noise variance minimization standpoint. A fusion-adaptive noise reduction process may further be performed on the resultant fused image.

Abstract: A safety system, for use with a power tool, includes a sensor assembly and a controller. The power tool includes a base and a moving component that is movable relative to the base. The sensor assembly monitors a predetermined danger zone that is adjacent to the moving component of the power tool. The sensor assembly is configured to generate data relating to the predetermined danger zone. The controller receives the data from the sensor assembly and analyzes the data from the sensor assembly to determine if at least a portion of a hand of the user is present within the predetermined danger zone. The safety system can further include a wearable component including infrared only reflective material that is coupled to the hand of the user. The controller analyzes the data from the sensor assembly to determine if the wearable component is present within the predetermined danger zone.

Abstract: A search device identifies names of POI from a document group having not been determined. A storage unit that stores a POI presence/absence learning model having learned contexts relating to presence/absence of POI, a POI state learning model having learned contexts relating to states of POI, and a POI name learning model having learned features relating to names of POI, an acceptance unit that accepts a first document group that is a determination target, first and second determination units and an identifying unit that identifies a name of a POI using the POI name learning model from each document of a third document group for which information relating to states of POI is determined by the second determination unit in a second document group are included.

Abstract: Systems and methods for processing video are provided. The method includes receiving a text-based description of active scenes and representing the text-based description as a word embedding matrix. The method includes using a text encoder implemented by neural network to output frame level textual representation and video level representation of the word embedding matrix. The method also includes generating, by a shared generator, frame by frame video based on the frame level textual representation, the video level representation and noise vectors. A frame level and a video level convolutional filter of a video discriminator are generated to classify frames and video of the frame by frame video as true or false. The method also includes training a conditional video generator that includes the text encoder, the video discriminator, and the shared generator in a generative adversarial network to convergence.

Abstract: Mechanisms for identification of text fields in documents using neural networks are described. Identification of text fields includes obtaining a plurality of symbol sequences of a document having a plurality of text fields, determining a plurality of vectors representative of one of the plurality of symbol sequences, processing the plurality of vectors using a first neural network to obtain, based on values of the plurality of vectors, a plurality of recalculated vectors, determining an association between a first recalculated vector of the plurality of recalculated vectors and a first text field of the plurality of text fields, the first recalculated vector being representative of a first symbol sequence of the plurality of symbol sequences, and determining, based on the association between the first recalculated vector and the first text field, an association between the first symbol sequence and the first text field.

Abstract: An example apparatus including memory to store a first image of a document and a second image of the document, and a processor coupled to the memory, wherein the processor is to: perform optical character recognition on the first image to generate a first output dataset; perform optical character recognition on the second image to generate a second output dataset; and compute a transformation matrix based on the first output dataset and the second output dataset, the transformation matrix to align the first image with the second image.

Abstract: Methods and systems for generating synthetic point cloud data are described. Projected 2D data grid is generated by projecting a 3D point cloud into a 2D grid, with rotation equivariance. A generative model is learned using the projected 2D data grid, wherein the generative model is implemented using flex-convolution and transpose flex convolution operations, for example in a generative adversarial network. The learned generative model is used to generate synthetic point clouds.

Abstract: A user proximity detection device including a receiver configured to receive a signal in a target frequency band, an inference-purpose data generator configured to measure an intensity of a signal received through the receiver and generate inference-purpose data based on the measured intensity of the signal, and a proximity detector configured to input the inference-purpose data into a human body proximity inference machine learning model to determine whether a human body is in proximity. The target frequency band is selected from a broadcast frequency band. In addition, at least one of an autonomous vehicle a user terminal or a server is linked or converged with an artificial intelligence module, an Unmanned Aerial Vehicle (UAV), a robot, an Augmented Reality (AR) device, Virtual Reality (VR), and devices related to 5G service.

Abstract: A method for measuring a hole provided in a workpiece is provided and the method comprises: obtaining a three-dimensional point cloud model of the workpiece and a two-dimensional image of the workpiece, defining a first contour in the three-dimensional point cloud model based on an intensity difference of the two-dimensional image, defining a second contour and a third contour respectively based in the first contour, bounding a data point testing region between the second contour and the third contour, respectively defining data point sampling regions along a plurality of cross-section directions of the data point testing region, respectively sampling data points in the data point sampling regions to obtain a turning point set comprising turning points, wherein each of the turning points has the largest turning margin, connecting the turning points which are distributed in the turning point set along a ring direction to obtain an edge of the hole.

Abstract: A method includes: executing a first generation process that includes generating a Betti number series corresponding to a contribution rate by performing persistent homology processing on a first point set, the first point set being generated by using a plurality of pieces of time series data and the contribution rate of each of the plurality of pieces of time series data, each of points included in the first point set being represented by coordinates; executing a second generation process that includes generating a characteristic image from a plurality of the Betti number series, the plurality of Betti number series being generated by performing the first generation process on each of the plurality of contribution rates; and executing a third generation process that includes generating machine learning data in which the characteristic image and a classification corresponding to the plurality of pieces of time series data are associated with each other.

Abstract: A method and apparatus for image processing. An image scaling circuit is configured to receive image data for a plurality of pixels corresponding to a first image. The image data includes color information and a transparency value for each of the plurality of pixels. The image scaling circuit updates the image data by selectively changing the color information for one or more of the pixels based at least in part on the transparency values. In updating the image data, the image scaling circuit may determine contextual information about the first image based at least in part on the transparency values. The image scaling circuit further generates an interpolated image based on the updated image data for the plurality of pixels. For example, the interpolated image may be a scaled version of the first image.

Abstract: An image processing method includes the following steps: receiving a two-dimensional image and segmenting an object block in the two-dimensional image, masking the object block with a color block to generate a mask image; inputting the two-dimensional image and the mask image into a first image processing model, outputting a feature vector from the first image processing model; and inputting the two-dimensional image into a second image processing model to obtain a feature map. The feature map comprises a plurality of feature channel maps, and the feature vector contains a plurality of feature values. Each of the feature channel maps corresponds to one of the feature values in sequence. A weighted feature map is generated according to the feature channel maps and the feature values.

Abstract: The present disclosure relates to an object selection system that accurately detects and automatically selects user-requested objects (e.g., query objects) in a digital image. For example, the object selection system builds and utilizes an object selection pipeline to determine which object detection neural network to utilize to detect a query object based on analyzing the object class of the query object. In addition, the object selection system can add, update, or replace portions of the object selection pipeline to improve overall accuracy and efficiency of automatic object selection within an image.

Abstract: A method and system for detecting defects in a specimen, the method comprising: obtaining an image comprising a plurality of pixels of a specimen part; processing the image according to a detection recipe to derive information related to potential defects in the specimen, the information comprising a first data set informative of first locations identified, in accordance with the detection recipe as locations of potential defects, and a second data set informative of second locations not identified as locations of potential defects; receiving data specifying targeted locations of interest within the part of the specimen; when the first data set is not informative of each targeted location, generating a third data set by adding to the first data set information related to the missing targeted location from the second data set, the information bearing an indication that it corresponds to a targeted location; and outputting the third data set.

Abstract: An electronic device and method are disclosed herein. The electronic device includes: a processor, and a memory operatively connected to the processor.

Abstract: An information processing method includes, by using a computer: acquiring an annotation skill of an operator who provides annotation information to image; acquiring annotation-target image having an image feature identified based on the annotation skill; acquiring an annotation result attained by the operator for the annotation-target image; and updating the annotation skill using the annotation result and reference annotation information corresponding to the image feature.