| 46th week of 2021 patent applcation highlights part 49 |
| Patent application number | Title | Published |
|---|
| 20210357644 | EXPLANATORY VISUALIZATIONS FOR OBJECT DETECTION - Introduced here are computer programs and associated computer-implemented techniques for creating visualizations to explain the outputs produced by models designed for object detection. To accomplish this, a graphics editing platform can obtain a reference output that identifies a region of pixels in a digital image that allegedly contains an object. Then, the graphics editing platform can compute the similarity between the reference output and a series of outputs generated by a model upon being applied to masked versions of the digital image. A visualization component can be produced based on the similarity. | 2021-11-18 |
| 20210357645 | COMPUTE SYSTEM WITH WEAR DETECTION MECHANISM AND METHOD OF OPERATION THEREOF - A method of operation of a compute system includes: capturing an image of a surface of a tire; identifying a tire wear in the image; categorizing the tire wear as a bald region, a crack, a foreign object, low tread, or a combination thereof; generating a wear report includes identifying the tire wear as the bald region, the crack, the foreign object, low tread, or the combination thereof; and transferring the wear report for displaying on a display. | 2021-11-18 |
| 20210357646 | Method and Computing Device in which Visual and Non-Visual Semantic Attributes are Associated with a Visual - The present invention provides a method in which visual and non-visual semantic attributes are associated with a visual comprising preferably an input step, a preliminary visual processing step, a semantic concept processing step, a semantic context processing step, a semantic marker processing step, a semantic inheritance processing step, a semantic instance processing step, and a lexical functions step, as well as a computing device which is capable of performing said method. | 2021-11-18 |
| 20210357647 | Method and System for Video Action Classification by Mixing 2D and 3D Features - A method, system, and computer program product provide for video action classification by selecting a first video frame and a first plurality of video frames from a received video to process the first video frame with a 2D convolutional neural network processing pathway to extract spatial features classifying the first video frame, and to process the first plurality of video frames with a 3D convolutional neural network processing pathway to extract spatiotemporal features classifying the first plurality of video frames so that the spatial features are combined with the spatiotemporal features to generate a classification label for the video action. | 2021-11-18 |
| 20210357648 | IMAGE PROCESSING NEURAL NETWORK SYSTEMS AND METHODS WITH SCENE UNDERSTANDING - An image processing neural network system includes a base net of at least one convolutional layer and at least one pooling layer; and a scenario block layer. The scenario block layer performs scene classification and generates a dictionary of scenarios and a vector of scenario encoding coefficients to output a probabilistic scene class assignment and the vector of scenario encoding coefficients. The vector of scenario encoding coefficients corresponds to reasoning for the scene classification. | 2021-11-18 |
| 20210357649 | SYSTEMS AND METHODS OF ENFORCING DISTANCING RULES - Disclosed herein are systems, methods, and non-transitory computer readable mediums directed to identifying persons monitored on video frames of a video feed, determining if the number of people on a monitored video frame is above a prescribed group threshold, and if so, calculating distances between each respective person in the monitored video frame. The calculated distances between each respective person are then compared to a prescribed distance threshold and alerts are generated in response to any calculated distance not satisfying the prescribed threshold. Identification of persons in a monitored video frame may be performed by, for example, facial recognition software. Calculation of distance between persons in a monitored video frame may be performed, for example, by using bounding boxes, machine learning, or other classification techniques. | 2021-11-18 |
| 20210357650 | BOWLING LANE ERROR DETECTION - An error detection system for notifying maintenance personnel of errors in a bowling lane operation may include a camera, artificial light source (optional), a computer that sends an output signal. The output signal may be a signal to provide notification to maintenance personnel to fix the error or an electrical signal to the pinsetter or ball return unit to manipulate operation of the same to prevent further damage. In this manner, interruptions and delays to bowlers will be minimized because maintenance personnel is immediately notified of any errors and can resolve the errors before the bowler realizes the error. | 2021-11-18 |
| 20210357651 | SYSTEMS AND APPROACHES FOR LEARNING EFFICIENT REPRESENTATIONS FOR VIDEO UNDERSTANDING - Systems and methods for performing video understanding and analysis. Sets of feature maps for high resolution images and low resolution images in a time sequence of images are combined into combined sets of feature maps each having N feature maps. A time sequence of temporally aggregated sets of feature maps is created for each combined set of feature maps by: selecting a selected combined set of feature maps corresponding to an image at time “t” in the time sequence of images; applying, by channel-wise multiplication, a feature map weighting vector to a number of combined sets of feature maps that are temporally adjacent to the selected combined set of feature maps; and summing elements of the number of combined set of feature maps into a temporally aggregated set of feature maps. The time sequence of temporally aggregated sets of feature maps is processed to perform video understanding processing. | 2021-11-18 |
| 20210357652 | METHOD, APPARATUS, ELECTRONIC DEVICE AND READABLE STORAGE MEDIUM FOR CLASSIFYING VIDEO - A method, an apparatus, an electronic device and a computer readable storage medium for classifying a video are provided. The method may include: acquiring a to-be-classified video and an image complexity of the to-be-classified video; extracting a feature map with a resolution corresponding to the image complexity from the to-be-classified video; and determining a target video class to which the to-be-classified video belongs based on the feature map. | 2021-11-18 |
| 20210357653 | METHOD AND APPARATUS FOR COMMENTING VIDEO - Embodiments of the present disclosure disclose a method and apparatus for commenting a video, and relate to the field of cloud computing. The method may include: acquiring content information of a to-be-processed video frame; constructing text description information based on the content information, the text description information being used to describe a content of the to-be-processed video frame; importing the text description information into a pre-trained text conversion model to obtain commentary text information corresponding to the text description information, the text conversion model being used to convert the text description information into the commentary text information; and converting the commentary text information into audio information. | 2021-11-18 |
| 20210357654 | SYSTEMS AND METHODS OF IDENTIFYING PERSONS-OF-INTEREST - Disclosed herein are systems, methods, and non-transitory computer readable mediums directed to tracing a person of interest (POI) using video. As provided herein, images of a POI are obtained, such as from a data store, and then the POI is identified in one or more video frames of a monitored video feed by comparing the monitored video frames to the images of the POI using facial recognition techniques. If the POI is identified, additional persons within a prescribed distance threshold of the POI are determined and identified using facial recognition techniques. Thereafter, a list of each identified person within the distance threshold of the POI is generated and may be transmitted to a desired recipient. | 2021-11-18 |
| 20210357655 | SHIP AND HARBOR MONITORING DEVICE AND METHOD - The present invention relates to a method by which a computing means monitors a harbor, and a harbor monitoring method, according to one aspect of the present invention, comprises the steps of: acquiring a harbor image; generating a segmentation image corresponding to the harbor image; generating a display image corresponding to the harbor image and having a first view attribute; generating a conversion segmentation image, which corresponds to the segmentation image and has a second view attribute different from the first view attribute; matching the display image so as to generate a panoramic image; matching the conversion segmentation image so as to generate a matching segmentation image; calculating ship mooring guide information on the basis of the matching segmentation image; and outputting the mooring guide information together with the panoramic image. | 2021-11-18 |
| 20210357656 | SYSTEMS AND METHODS FOR OBJECT RECOGNITION - The present disclosure relates to systems and methods for object recognition. The system may obtain an image and a model. The image may include a search region in which the object recognition process is performed. In the objection recognition process, for each of one or more sub-regions of the search region, the system may determine a match metric indicating a similarity between the model and the sub-region of the search region. Further, the system may determine an instance of the model among the one or more sub-regions of the search region based on the match metrics. | 2021-11-18 |
| 20210357657 | METHODS, SYSTEMS, APPARATUSES, AND DEVICES FOR FACILITATING MANAGING INCIDENTS OCCURRING IN AREAS MONITORED BY LOW DATA-RATE MONITORING DEVICES USING THE LOW DATA-RATE MONITORING DEVICES - Disclosed herein is a system for facilitating managing incidents occurring in areas monitored by low data-rate monitoring devices using the low data-rate monitoring devices, in accordance with some embodiments. Accordingly, the system comprises a processor, a device server, and a data visualization device. Further, a camera of a low data-rate monitoring device is configured for capturing video of an area. Further, the processor comprises a machine learning (ML) hardware accelerator configured for performing machine learning processing of the video. Further, the processor is configured for generating processed data. Further, the device server is configured for receiving the processed data based on the transmitting of the processed data from a low data-rate transceiver of the low data-rate monitoring device and transmitting a notification to a device. Further, the data visualization device is configured for visualizing the processed data, identifying an incident in the area, and generating the notification for the incident. | 2021-11-18 |
| 20210357658 | SYSTEM AND METHOD FOR DETECTING SCAN IRREGULARITIES AT SELF-CHECKOUT TERMINALS - A system for detecting a scan irregularity in scanning process during check-out at a retail store, includes an image receiving module for receiving a video stream of a scanning zone, an image processing module for detecting visual scan intervals in image frames of the video stream, and a decision module. The decision module is configured to process each detected visual scan interval, wherein a processed visual scan interval includes a valid scan action, wherein the valid scan action is a user action performed for scanning an item. The decision module is further configured to detect a scan irregularity in the check-out process, wherein the scan irregularity occurs when an item identified for scanning in a processed visual scan interval is absent in a list of scanned items generated by the scanner during corresponding interval, and provide an alert regarding the scan irregularity at a user computing device. | 2021-11-18 |
| 20210357659 | OBJECT TRACKING APPARATUS, OBJECT TRACKING SYSTEM, OBJECT TRACKING METHOD, DISPLAY CONTROL DEVICE, OBJECT DETECTION DEVICE, AND COMPUTER-READABLE MEDIUM - An object tracking apparatus, method and computer-readable medium for detecting an object from output information of sensors, tracking the object on a basis of a plurality of detection results, generating tracking information of the object represented in a common coordinate system, outputting the tracking information, and detecting the object on a basis of the tracking information. | 2021-11-18 |
| 20210357660 | METHOD, APPARATUS, ELECTRONIC DEVICE, AND STORAGE MEDIUM FOR MONITORING AN IMAGE ACQUISITION DEVICE - A method, an apparatus, an electronic device, and a storage medium for monitoring an image acquisition device are provided, which are related to a field of computer vision technology. The method for monitoring the image acquisition device includes: determining a stop position of each target vehicle from a first video image acquired by the image acquisition device; determining a marking line in the first video image according to the stop positions of their respective target vehicles, and determining a deviation amount of the marking line in the first video image from a reference making line, and determining that the image acquisition device has deviated in a case that the deviation amount reaches a predetermined condition. The workload of manual detection is reduced, and the accuracy of deviation amount monitoring is increased by comparing the marking line with the reference marking line. | 2021-11-18 |
| 20210357661 | A METHOD OF TRACKING OBJECTS IN A SCENE - A method of detecting objects located in an environment around a vehicle, comprises the steps of obtaining a 2D image of a scene from an image capture device fitted to the vehicle, generating a set of particles that each comprise a set of parameters that define the state of at least one object that may be present in the scene, calculating an edge strength and edge direction for each pixel of the captured 2D image; and for each particle in the set: generating a 3D probabilistic model of the object that is associated with the particle, the model defining a set of edges of the object, each edge in the model defined by at least one likelihood function defining a probabilistic distribution of at least one characteristic of the edge, mapping the likelihood functions of the probabilistic model into the 2D plane of the captured image as a function of the parameters of the particle; and processing the edge strength and edge direction values with the probability values to determine the likelihood that the particle defines the state of an object present in the environment. | 2021-11-18 |
| 20210357662 | GROUND TRUTH BASED METRICS FOR EVALUATION OF MACHINE LEARNING BASED MODELS FOR PREDICTING ATTRIBUTES OF TRAFFIC ENTITIES FOR NAVIGATING AUTONOMOUS VEHICLES - A system uses a machine learning based model to determine attributes describing states of mind and behavior of traffic entities in video frames captured by an autonomous vehicle. The system classifies video frames according to traffic scenarios depicted, where each scenario is associated with a filter based on vehicle attributes, traffic attributes, and road attributes. The system identifies a set of video frames associated with ground truth scenarios for validating the accuracy of the machine learning based model and predicts attributes of traffic entities in the video frames. The system analyzes video frames captured after the set of video frames to determine actual attributes of the traffic entities. Based on a comparison of the predicted attributes and actual attributes, the system determines a likelihood of the machine learning based model making accurate predictions and uses the likelihood to generate a navigation action table for controlling the autonomous vehicle. | 2021-11-18 |
| 20210357663 | ROAD RECOGNITION DEVICE - A road recognition device includes a surroundings recognition section recognizing, as surroundings information, at least one of a shape of a roadside object and a travel history of another vehicle, a reliability setting section setting reliability of the surroundings information, a reference line setting section preferentially using surroundings information having higher reliability to determine a reference line of an own lane, and an output section outputting the reference line. When a direction indicator is in operation, the reliability setting section sets reliability of the surroundings information for a direction opposite to a direction indicated by the direction indicator so as to be lower. When the direction indicator is in operation, and the vehicle is traveling in a lane-change prohibition section, the reliability setting section sets reliability of the surroundings information including at least one of the shape of the roadside object and the travel history so as to be lower. | 2021-11-18 |
| 20210357664 | OBSTACLE MONITORING SYSTEMS AND METHODS FOR SAME - An autonomous obstacle monitoring and vehicle control system includes a remote sensing device including one or more sensors. The remote sensing device is movable relative to an agricultural system, and configured to observe obstacles proximate to a path of an agricultural system or proximate to the agricultural system. An obstacle recognition module communicates with the remote sensing device, and is configured to identify and index obstacles proximate to the path or proximate to the agricultural system. An autonomous agricultural system controller is configured for communication with the agricultural system. The autonomous agricultural system controller includes a mission administration module configured to operate the remote sensing device, and a vehicle operation module configured to control the agricultural system based on the identified and indexed obstacles. | 2021-11-18 |
| 20210357665 | SAFETY SYSTEM FOR A VEHICLE TO DETECT AND WARN OF A POTENTIAL COLLISION - Systems and methods are provided for processing reports received from vehicles. A processing device perform operations comprising receiving a first report from a first vehicle, the first report generated by the first vehicle for a first hazard detected by the first vehicle; receiving a second report from a second vehicle, the second report generated by the second vehicle for a second hazard detected by the second vehicle; analyzing the first report and the second report to make a determination that the first report and the second report identify a related hazard; aggregating the first report and the second report into a consolidated report based on the determination that the first report and the second report identify a related hazard; and processing the consolidated report to identify a contributing factor to the related hazard. | 2021-11-18 |
| 20210357666 | IMAGE PROCESSOR AND IMAGE PROCESSING METHOD - An image processor includes a boundary line detection portion configured to detect a boundary line by using an image in accordance with an image signal acquired by an imaging device, a parking frame detection portion configured to detect a parking frame by using the detected boundary line, an in-frame scanning portion configured to acquire a dividing point that divides a pair of facing sides of the boundary lines that define the parking frame and to scan the image through the dividing point to detect an edge, a storage portion configured to store a state of an edge and an attribute of the parking frame by linking them, and a determination portion configured to determine an attribute of the parking frame by using a state of the detected edge and the stored attribute of the parking frame that is linked to the state of the edge stored in the storage portion. | 2021-11-18 |
| 20210357667 | Methods and Systems for Measuring and Mapping Traffic Signals - Disclosed herein are methods and systems for measuring and mapping traffic signals in global coordinates. A method includes obtaining an image of an environment for a vehicle location, detecting a traffic signal in the image, classifying the detected traffic signal, obtaining traffic signal specification information associated with the classified traffic signal, determining a control point associated with the classified traffic signal, determining a 3D position of the control point in camera space using the traffic signal specification information, camera calibration information, image pixel size, and focal length of camera used in capturing the image, transforming the 3D position of the control point in the camera space to a 3D position of the control point in global coordinates, saving the 3D position of the control point in global coordinates in a map, and controlling operation of a vehicle with 3D positions of control points in global coordinates saved in the map. | 2021-11-18 |
| 20210357668 | DIFFERENTIATION-BASED TRAFFIC LIGHT DETECTION - A method, apparatus, and system for determining a state of an upcoming traffic light is disclosed. At an autonomous driving vehicle (ADV), an upcoming traffic light ahead in a direction of travel is detected. A relative position of the ADV to the traffic light is determined based on a three-dimensional ( | 2021-11-18 |
| 20210357669 | METHOD FOR CONTROLLING AUTONOMOUS VEHICLE - A method for controlling an autonomous vehicle is disclosed. The vehicle control method, which adjusts a seat installed in a vehicle, a first display positioned in front of the seat and facing the seat, and a second display positioned in front of the first display and facing the seat, includes: detecting the face of a passenger sitting in the seat by a camera installed in the vehicle; estimating the face height from the floor surface of the vehicle; if the estimated face height is lower than a predetermined level, setting a display region on the first display, and, if the estimated face height is higher than the predetermined level, setting the display region on at least part of the second display; and displaying images in the set display region. One or more among an autonomous vehicle, user terminal, and server according to the present invention may be associated with an artificial intelligence module, a robot, an augmented reality (AR) device, a virtual reality (VR) device, etc. | 2021-11-18 |
| 20210357670 | Driver Attention Detection Method - The disclosure relates to technology for monitoring driver attentiveness in a vehicle. A driver distraction system collects vehicle data and scene information from the vehicle while traveling on a route. The vehicle data and scene information are then processed to generate a reference heat map. At the same time, the driver distraction system may capture a gaze of a driver to track a gaze direction and duration of the driver while driving the vehicle on the route. The gaze direction and duration are processed to generate a driver gaze heat map. The driver gaze heat map and reference heat map are analyzed to determine a level of driver distraction of the driver in the vehicle, and a recommendation or warning is output to the driver of the vehicle according to the level of driver distraction. | 2021-11-18 |
| 20210357671 | SPOOF DETECTION USING IRIS IMAGES - The technology described in this document can be embodied in a method for preventing access to a secure system based on determining a captured image to be of an alternative representation of a live person. The method includes capturing an image of a subject illuminated by an infrared (IR) illumination source, and extracting, from the image, a portion representative of an iris of the subject. The method also includes determining that an amount of high-frequency features in the portion of the image satisfies a threshold condition indicative of the image being of an alternative representation of a live person, and in response, identifying the subject in the image to be an alternative representation of a live person. Responsive to identifying the subject in the image to be an alternative representation of a live person, the method further includes preventing access to the secure system. | 2021-11-18 |
| 20210357672 | DISPLAY SUBSTRATE, METHOD FOR MANUFACTURING THE SAME AND DISPLAY DEVICE - A display substrate, a method for manufacturing the same and a display device are provided. The display substrate includes a base substrate and a thin film transistor array arranged on the base substrate. Multiple pixels arranged in an array are provided in an effective display region of the display substrate. The effective display region includes an optical element arrangement region and other display regions, and a transmittance of the optical element arrangement region is larger than transmittances of the other display regions. In the optical element arrangement region, an optical element is arranged on a side of the base substrate away from the thin film transistor array, and the optical element emits and receives light that is transmitted through the display substrate. | 2021-11-18 |
| 20210357673 | METHOD AND DEVICE FOR PROCESSING FEATURE POINT OF IMAGE - A method and a device for processing feature points of an image are provided. A specific embodiment of the method includes obtaining an image to be processed; determining weights of the feature points of the image to be processed to obtain a weight set; and according to the weights, selecting target numbered feature points as target feature points of the image to be processed. The weights include a texture weight; the texture weight and a color change scope of pixels in a target sized image region in which the feature points locate are directly proportional. The embodiment can reduce the number of feature points of the image, and further release the storage pressure of feature points regarding the image. | 2021-11-18 |
| 20210357674 | IMAGE PROCESSING SYSTEM, IMAGE PROCESSING METHOD, AND STORAGE MEDIUM EACH FOR OBTAINING PIXELS OF OBJECT USING NEURAL NETWORK - A first decoder obtains a region including an object targeted for recognition based on a feature map obtained by performing convolutional processing on a processing target image. Next, the first decoder obtains, from the feature map, a partial feature map of a portion corresponding to the obtained region including the object targeted for recognition. Then, a second decoder obtains pixels corresponding to the object targeted for recognition based on the partial feature map. This reduces the amount of calculation required for decoder units included in a neural network. | 2021-11-18 |
| 20210357675 | IMAGE PROCESSING APPARATUS, IMAGE PICKUP APPARATUS, IMAGE PROCESSING METHOD, AND STORAGE MEDIUM - An image processing apparatus includes a separation unit configured to separate luminance information of each pixel in a processing area of an input image into an intrusion component and an object component, a determination unit configured to determine whether each of intrusion components is a first component or a second component, a processing unit configured to generate first information based on the first component, and an image generating unit configured to generate an output image based on the first information, the second component, and the object component. | 2021-11-18 |
| 20210357676 | INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND STORAGE MEDIUM - An information processing apparatus determining whether image capturing by an image capturing apparatus is obstructed, the information processing apparatus comprising: a dividing unit configured to divide an input image captured by the image capturing apparatus into a plurality of blocks; a processing determination unit configured to determine whether to perform first detection processing using a reference image corresponding to the image capturing apparatus or second detection processing using a feature of the input image, on each of the blocks; and an obstruction determination unit configured to determine whether the image capturing by the image capturing apparatus is obstructed, based on a detection result of each of the blocks by the first detection processing or the second detection processing. | 2021-11-18 |
| 20210357677 | CLASSIFICATION SYSTEM AND LEARNING SYSTEM - A classification system according to an embodiment includes a score calculation unit, a determination unit, and a classification unit. The score calculation unit calculates respective scores of predetermined classes from input data. The determination unit determines whether the input data belongs to anyone of the classes based on the respective scores of the classes, which are calculated by the score calculation unit. The classification unit determines which one of the classes the input data belongs to, based on the calculated scores when the determination unit determines that the input data belongs to anyone of the classes and determines that the input data belongs to an unknown class that is other than the classes when the determination unit determines that the input data does not belong the classes. | 2021-11-18 |
| 20210357678 | INFORMATION PROCESSING METHOD AND APPARATUS, AND STORAGE MEDIUM - An information processing method and apparatus, and a storage medium, the method including: on the basis of aggregated archive data, determining a target object; acquiring first capture image information of the target object; analysing the first capture information to obtain a first analysis result; and, on the basis of the first analysis result, determining a first trajectory of the target object. | 2021-11-18 |
| 20210357679 | CLUSTERING TECHNIQUES FOR MACHINE LEARNING MODELS - In some aspects, systems and methods for efficiently clustering a large-scale dataset for improving the construction and training of machine-learning models, such as neural network models, are provided. A dataset used for training a neural network model configured can be clustered into a first set of clusters and a second set of clusters. The neural network model can be constructed with a number of nodes in a hidden layer that is based on the number of clusters in the first set of clusters. The neural network can be trained based on training samples selected from the second set of clusters. In some aspects, the trained neural network model can be utilized to satisfy risk assessment queries to compute output risk indicators for target entities. The output risk indicator can be used to control access to one or more interactive computing environments by the target entities. | 2021-11-18 |
| 20210357680 | MACHINE LEARNING CLASSIFICATION SYSTEM - A computing device classifies unclassified observations. A first batch of unclassified observation vectors and a first batch of classified observation vectors are selected. A prior regularization error value and a decoder reconstruction error value are computed. A first batch of noise observation vectors is generated. An evidence lower bound (ELBO) value is computed. A gradient of an encoder neural network model is computed, and the ELBO value is updated. A decoder neural network model and an encoder neural network model are updated. The decoder neural network model is trained. The target variable value is determined for each observation vector of the unclassified observation vectors based on an output of the trained decoder neural network model. The target variable value is output. | 2021-11-18 |
| 20210357681 | Scalable Attributed Graph Embedding for Large-Scale Graph Analytics - A computer-implemented method for calculating Scalable Attributed Graph Embedding for Large-Scale Graph Analytics that includes computing a node embedding for a first node-attributed graph in a node embedded space. One or more random attributed graphs is generated in the node embedded space. A graph embedding operation is performed using a dissimilarity measure between one or more raw graphs and the one or more generated random graphs, and an edge-attributed graph into a second node-attributed graph using an adjoint graph. | 2021-11-18 |
| 20210357682 | ARTIFICIAL INTELLIGENCE DRIVEN IMAGE RETRIEVAL - A method for retrieving a plurality of electronic images that includes obtaining images from a plurality of electronic storage sites. A first set of images containing a main protagonist is selected from the plurality of electronic images. An intent based image selection is performed from the first set of images. The intent based image selection includes tagging the first set of images with word tags identifying content of the images, creating a word cloud from the word tags, plotting the top word tags from the word cloud in a Venn diagram, and extracting images from the overlapping are of the Venn diagram. | 2021-11-18 |
| 20210357683 | METHOD AND APPARATUS FOR DETERMINING TARGET ANCHOR, DEVICE AND STORAGE MEDIUM - Embodiments of the present disclosure disclose a method and apparatus for determining a target anchor, a device and a storage medium. The method may include: extracting a plurality of feature maps of an original image using a feature extraction network; inputting the plurality of feature maps into a feature pyramid network to perform feature fusion, to obtain a plurality of fused feature maps; and using a region proposal network to implement operations as follows: determining an initial anchor of a web header using the fused feature map, based on a size of each fused feature map, and determining an offset parameter of the initial anchor, based on a ratio of the size of the fused feature map to the original image, and generating a plurality of candidate anchors in different directions, based on the offset parameter of the initial anchor. | 2021-11-18 |
| 20210357684 | Labeling Techniques for a Modified Panoptic Labeling Neural Network - A panoptic labeling system includes a modified panoptic labeling neural network (“modified PLNN”) that is trained to generate labels for pixels in an input image. The panoptic labeling system generates modified training images by combining training images with mask instances from annotated images. The modified PLNN determines a set of labels representing categories of objects depicted in the modified training images. The modified PLNN also determines a subset of the labels representing categories of objects depicted in the input image. For each mask pixel in a modified training image, the modified PLNN calculates a probability indicating whether the mask pixel has the same label as an object pixel. The modified PLNN generates a mask label for each mask pixel, based on the probability. The panoptic labeling system provides the mask label to, for example, a digital graphics editing system that uses the labels to complete an infill operation. | 2021-11-18 |
| 20210357685 | AUTOMATED PART-INFORMATION GATHERING AND TRACKING - A method for identifying a component part includes receiving a digital image of an object and textual information about the object and accessing images of component parts and textual information about the component parts. The method further includes applying the digital image to a first classifier trained on the images of the component parts to classify the object as a first of the component parts and applying the textual information about the object to a second classifier trained on the textual information about the component parts to recognize the textual information as information about the first of the component parts or a second of the component parts. The method further includes identifying the object as a component part that is the first of the component parts or the second of the component parts and accessing a data record with information about the component part. | 2021-11-18 |
| 20210357686 | FINETUNE IMAGE FEATURE EXTRACTION USING ENVIRONMENTAL DATA - A method, system, and computer program product for determining selection parameters for filtering algorithms using environmental data of images. The method may include receiving an image. The method may also include analyzing the image using at least image processing. The method may also include identifying, based on the analyzing, image data and environmental data of the image. The method may also include inputting the image data and the environmental data into a machine learning algorithm, where the machine learning algorithm includes mapped relationships between at least the environmental data and selection parameters. The method may also include predicting, using the machine learning algorithm, optimal selection parameters for the image. The method may also include applying the optimal selection parameters to a filtering algorithm for the image. The system and computer program product may include similar steps. | 2021-11-18 |
| 20210357687 | SYSTEMS AND METHODS FOR PARTIALLY SUPERVISED ONLINE ACTION DETECTION IN UNTRIMMED VIDEOS - Embodiments described herein provide systems and methods for a partially supervised training model for online action detection. Specifically, the online action detection framework may include two modules that are trained jointly—a Temporal Proposal Generator (TPG) and an Online Action Recognizer (OAR). In the training phase, OAR performs both online per-frame action recognition and start point detection. At the same time, TPG generates class-wise temporal action proposals serving as noisy supervisions for OAR. TPG is then optimized with the video-level annotations. In this way, the online action detection framework can be trained with video-category labels only without pre-annotated segment-level boundary labels. | 2021-11-18 |
| 20210357688 | Artificial Intelligence System For Automated Extraction And Processing Of Dental Claim Forms - A dental form image may be processed with a segmentation network to identify point labels corresponding to reference point labels of a reference form. The image and the point labels along with a reference image and the reference point labels may be processed by a pair of encoders to obtain offsets. Text blobs may be identified from portions of the image corresponding to the reference point labels, such as with correction according to the offsets. Image portions and text blobs for each field of the dental form may be processed to extract text. Intermediate values of machine learning models used to extract text may be input to a machine learning model estimating a procedure code for the dental form. Machine learning models may be used to correctly identify a provider referenced by the dental form. | 2021-11-18 |
| 20210357689 | COMPUTER-IMPLEMENTED METHOD AND SYSTEM FOR TRAINING AN EVALUATION ALGORITHM, COMPUTER PROGRAM AND ELECTRONICALLY READABLE DATA CARRIER - A computer-implemented method for the further training of an artificial-intelligence evaluation algorithm that has already been trained based upon basic training data, wherein the evaluation algorithm ascertains output data describing evaluation results from input data comprising image data recorded with a respective medical imaging facility. In an embodiment, the method includes ascertaining at least one additional training data set containing training input data and training output data assigned thereto; and training the evaluation algorithm using the at least one additional training data set. The additional training data set is ascertained from the input data used during a clinical examination process with a medical imaging facility, which the already-trained evaluation algorithm was used, and output data of the already-trained evaluation algorithm that has been at least partially correctively modified by the user. | 2021-11-18 |
| 20210357690 | ENHANCED NEURAL NETWORK SYSTEMS AND METHODS - Two stages of a convolutional neural network are linked by an interconnect that effects a spatial transposition of array data. The spatial transposition can include rotation, scaling, or translation (e.g., in x- or y-directions). A parameter characterizing the transposition (e.g., a parameter identifying rotation angle) can be learned by the same training process that is also used to learn other network parameters, such as layer coefficients. Additionally, or alternatively, data input to a neural network comprises—for each pixel in a patch of imagery—plural data that each indicates a relationship between the value of the pixel, and the value of a neighboring pixel. Some such neural networks can be trained to indicate the presence of a digital watermark signal in the patch of imagery—or a parameter characterizing such a digital watermark signal. Other features and arrangements are also detailed. | 2021-11-18 |
| 20210357691 | METHODS AND APPARATUS FOR PERFORMING ANALYTICS ON IMAGE DATA - Methods and apparatus for applying data analytics such as deep learning algorithms to sensor data. In one embodiment, an electronic device such as a camera apparatus including a deep learning accelerator (DLA) communicative with an image sensor is disclosed, the camera apparatus configured to evaluate unprocessed sensor data from the image sensor using the DLA. In one variant, the camera apparatus provides sensor data directly to the DLA, bypassing image signal processing in order to improve the effectiveness the DLA, obtain DLA results more quickly than using conventional methods, and further allow the camera apparatus to conserve power. | 2021-11-18 |
| 20210357692 | MULTI-FIDELITY SIMULATED DATA FOR MACHINE LEARNING - A method of training a machine learning system. The method comprises collecting a first simulation dataset derived from a computer simulating a hypothetical scenario with a first simulation configuration having a first degree of fidelity. The method further comprises collecting a second simulation dataset derived from a computer simulating the hypothetical scenario with a second simulation configuration having a second degree of fidelity different than the first degree of fidelity. The method further comprises building a multi-fidelity training dataset including training data from both the first simulation dataset and the second simulation dataset according to an interleaving protocol. | 2021-11-18 |
| 20210357693 | SUBSTRATE INSPECTION APPARATUS AND METHOD OF DETERMINING FAULT TYPE OF SCREEN PRINTER - A substrate inspection apparatus generates, when anomalies of a plurality of second solder pastes among a plurality of first solder pastes printed on a first substrate is detected, at least one image indicating a plurality of second solder pastes with anomalies detected by using an image about a first substrate, applies the at least one image to a machine-learning-based model, acquires a plurality of first values indicating relevance of respective first fault types to the at least one image and a plurality of first images indicating regions associated with one of a plurality of first fault types, determines a plurality of second fault types, which are associated with the plurality of second solder pastes by using the plurality of first values and the plurality of first images, and determines at least one third solder paste, which is associated with the respective second fault types. | 2021-11-18 |
| 20210357694 | LEARNING APPARATUS AND LEARNING METHOD FOR THREE-DIMENSIONAL IMAGE - A 3D image sliced into a plurality of slices including the first slice on which a label is annotated and a plurality of second slices on which the label is not annotated is provided as a training sample. A computing device trains a neural network based on the first slice, determines an expandable second slice which is expandable from the first slice from among the plurality of second slices based on the trained neural network; and trains the neural network based on expanded slices including the expandable second slice. | 2021-11-18 |
| 20210357695 | DEVICE AND METHOD FOR SUPPORTING GENERATION OF LEARNING DATASET - A learning dataset generation support device | 2021-11-18 |
| 20210357696 | PROCESSING FUNDUS CAMERA IMAGES USING MACHINE LEARNING MODELS TRAINED USING OTHER MODALITIES - Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a fundus image processing machine learning models that is configured to process one or more fundus images captured by a fundus camera to generate a predicted label. One of the methods includes generating training data, comprising: receiving sets of one or more training fundus images captured by a fundus camera; receiving, for each of the sets, a ground truth label assigned to a different image of the eye of the patient corresponding to the set that has been captured using a different imaging modality; and generating, for each set of training fundus images, a training example that includes the set of training fundus images in association with the ground truth label assigned to the different image of the patients eye; and training the machine learning model on the training examples in the training data. | 2021-11-18 |
| 20210357697 | TECHNIQUES TO EMBED A DATA OBJECT INTO A MULTIDIMENSIONAL FRAME - Various embodiments are generally directed to techniques for embedding a data object into a multidimensional frame, such as for training an autoencoder to generate latent space representations of the data object based on the multidimensional frame, for instance. Additionally, in one or more embodiments latent space representations of data objects may be classified, such as with a machine learning algorithm. Some embodiments are particularly directed to embedding a data object comprising a plurality of object entries into a three-dimensional (3D) frame. | 2021-11-18 |
| 20210357698 | IMAGE CLASSIFIER LEARNING DEVICE, IMAGE CLASSIFIER LEARNING METHOD, AND PROGRAM - An object is to make it possible to train an image recognizer by efficiently using training data that does not include label information. A determination unit | 2021-11-18 |
| 20210357699 | DATA QUALITY ASSESSMENT FOR DATA ANALYTICS - The invention relates to an approach for data quality assessment for data analytics, the approach comprising providing a data set, the data set comprising multiple data fields, predicting by a first trained machine learning model at least one usage type of the data set using characteristics of the data fields as input, for each usage type of the at least one usage type, determining a usage specific data quality score of each of the predicted usage types, and using of the data set based on the at least one usage type and associated data quality score. | 2021-11-18 |
| 20210357700 | METHOD AND APPARATUS FOR IMAGE ANALYSIS USING IMAGE CLASSIFICATION MODEL - A method for image analysis according to an embodiment may include generating a prediction result for an original image using a pre-trained image classification model, learning a plurality of masks using the original image, the prediction result, and the image classification model, and generating a map visualizing a importance of each area of the original image for the prediction result based on at least one of the plurality of masks. | 2021-11-18 |
| 20210357701 | EVALUATION DEVICE, ACTION CONTROL DEVICE, EVALUATION METHOD, AND EVALUATION PROGRAM - An evaluation device is provided and includes: a data acquisition unit which acquires input data input to a first learner of finishing learning that has undergone supervised learning using a data set including a pair of training data and correct answer data that indicates the correct answer of an inference result to the training data; and a validity evaluation unit which evaluates, based on an output obtained from a second learner of finishing learning that has undergone unsupervised learning using the training data by inputting the input data to the second learner, whether the valid output is able to be obtained from the first learner as the inference result to the input data when the input data is input to the first learner, the learning-finish second learner. | 2021-11-18 |
| 20210357702 | SYSTEMS AND METHODS FOR STATE IDENTIFICATION AND CLASSIFICATION OF TEXT DATA - The present disclosure provides systems and methods for identifying one or more states of a text string describing an event and classifying the event based on the one or more identified states. A method of this disclosure comprises receiving a text string describing an event, transforming the text string into modellable data, analyzing the word composition in the transformed data to identify one or more states of the event, and classifying the event based on the identified states. | 2021-11-18 |
| 20210357703 | SYSTEMS AND METHODS FOR FAULT CLASSIFICATION IN PHOTOVOLTAIC ARRAYS USING GRAPH SIGNAL PROCESSING - Various embodiments of a system and associated method for detecting and classifying faults in a photovoltaic array using graph-based signal processing. | 2021-11-18 |
| 20210357704 | SEMI-SUPERVISED LEARNING WITH GROUP CONSTRAINTS - A computer-implemented method for classification of data by a machine learning system using a logic constraint for reducing a data labeling requirement. The computer-implemented method includes: generating a first embedding space from a first partially labeled training data set, wherein in the first embedding space, content-wise related training data of the first partially labeled training data are clustered together, determining at least two clusters in the first embedding space formed from the first partially labeled training data, and training a machine learning model based, at least in part, on a second partially labeled training data set and the at least two clusters, wherein the at least two clusters are used as training constraints. | 2021-11-18 |
| 20210357705 | METHOD AND APPARATUS FOR CLASSIFYING IMAGE - An apparatus for classifying an image according to an embodiment includes a fake image generation module receiving a classification target image and a fake image in a form in which only a background exists and no specific object exists in the classification target image, a difference of images vector generation module generating a difference of images between the classification target image and the fake image and a difference of images vector by converting the generated difference of images into preset one-dimensional matrix data, a difference of feature vectors generation module generating a difference of feature vectors between a feature vector generated based on the classification target image and a feature vector generated based on the fake image, and an image classification module classifying the classification target image based on the difference of images vector, the feature vector generated based on the classification target image, and the difference of feature vectors. | 2021-11-18 |
| 20210357706 | SYSTEMS AND/OR METHODS FOR MACHINE-LEARNING BASED DATA CORRECTION AND COMPLETION IN SPARSE DATASETS - Certain example embodiments herein relate to techniques for automatically correcting and completing data in sparse datasets. Records in the dataset are divided into groups with properties having similar values. For each group, one or more properties of the records therein that is/are to be ignored is/are identified, based on record distances relative to the records in the group, and distances among values for each of the properties of the records in the respective group. The records in the groups are further divided into sub-groups without regard to the one or more properties that is/are to be ignored. The sub-groups include a smaller and more cohesive set of records. For each sub-group: based on the records therein, predicted values to be applied to values identified as being empty but needing to be filled in are determined; and those predicted values are applied. The corrected/completed dataset is provided as output. | 2021-11-18 |
| 20210357707 | VERIFICATION OF ELECTRONIC IDENTITY COMPONENTS - A unit-classification system receives a data set with identity data objects corresponding to personal identity components. Feature vectors are determined for the identity data objects. A trained classifier model determines, based on a feature vector for each identity data object, whether the corresponding personal identity components are included in a identity component (“IC”) category. The unit-classification system generates an IC identification for a first IC category, and associates the IC identification with a first identity data object corresponding to a first personal identity component. The unit-classification system identifies a second identity data object corresponding to a second personal identity component included in the first IC category. The unit-classification system modifies the first and second identity data objects to include the IC identification. Responsive to a request for the IC identification, the unit-classification system can provide a response that indicates the modified first and second identity data objects. | 2021-11-18 |
| 20210357708 | OBJECT DETECTION DEVICE, OBJECT DETECTION METHOD, AND PROGRAM - An object detection device detects a predetermined object from an image. The object detection device includes a first detection unit configured to detect a plurality of candidate regions where the predetermined object exists from the image, a region integrating unit configured to determine one or a plurality of integrated regions according to the plurality of candidate regions detected by the first detection unit, and a second detection unit configured to detect, in the one or the plurality of integrated regions, the predetermined object by using a detection algorithm different from an algorithm of the first detection unit. As a result, it is possible to detect the predetermined object faster and more accurately than before. | 2021-11-18 |
| 20210357709 | OBJECT DETERMINATION APPARATUS - In an object determination apparatus, a same-object determiner is configured to make a same-object determination as to whether a first object ahead of a subject vehicle that is a vehicle carrying the object determination apparatus, detected by an electromagnetic wave sensor, and a second object ahead of the subject vehicle, detected by an image sensor, are the same object. A candidate-object identifier is configured to identify a candidate for the first object, between which and the second object the same-object determination is to be made, as a candidate object. A candidate-object selector is configured to, in response to there being a plurality of the candidate objects, preferentially select, from the plurality of candidate objects, a candidate object whose likelihood for the identified object type of the second object is higher than a predetermined likelihood threshold, as a candidate object to be subjected to the same-object determination. | 2021-11-18 |
| 20210357710 | TEXT RECOGNITION METHOD AND DEVICE, AND ELECTRONIC DEVICE - A text recognition method includes: acquiring an image including text information, the text information including M characters, M being a positive integer greater than 1; performing text recognition on the image to acquire character information about the M characters; recognizing reading direction information about each character in accordance with the character information about the M characters, the reading direction information being used to indicate a next character corresponding to a current character in a semantic reading order; and ranking the M characters in accordance with the reading direction information about the M characters to acquire a text recognition result of the text information. | 2021-11-18 |
| 20210357711 | IMAGE FORMING APPARATUS AND IMAGE FORMING METHOD - An image forming apparatus includes a plurality of print heads, a sensor, a processor, and an image forming unit. The plurality of print heads are arranged in parallel. The sensor detects a color shift amount of each color to be printed by the plurality of print heads in a main scanning direction based on an output result of a test pattern by the plurality of print heads. The processor sets a color shift correction value for correcting the color shift amount on the basis of a color having a largest color shift amount. The processor controls light emission of the plurality of print heads based on the color shift correction value and image data. The image forming unit forms an image based on the image data on a sheet by light emission of the plurality of print heads. | 2021-11-18 |
| 20210357712 | INFORMATION PROCESSING DEVICE OUTPUTTING PRECEDING OPERATION COMMAND TO PRINTER - In an information processing device, a support program supports a printer, and a printing program is built in an operating system. The support program causes the information processing device to perform: in a case that a print instruction to execute printing by using the built-in printing program is issued, at least one of a command outputting process and an outputting instruction process. The command outputting process outputs a preceding operation command to the printer before starting transmitting print execution data to the printer. The preceding operation command commands the printer to execute a preceding operation prior to starting printing. The printer having a function to execute the preceding operation specified in the preceding operation command The outputting instruction process instructs a command transmission program to output the preceding operation command while the outputting instruction process designates the printer as an outputting destination. | 2021-11-18 |
| 20210357713 | IMAGE PROCESSING APPARATUS - An image processing apparatus is equipped with: an image data receiving unit that receives image data; and a print data generating unit that generates multiple value data as print data by performing a halftone process on the image data and converting the image data into at least three values. The print data generating unit administers a halftone process that includes a density fluctuation suppressing process that results in the print data always including a zero value except for a case in which all of the pixels that constitute the image data are converted into a maximum value. | 2021-11-18 |
| 20210357714 | PRINTING APPARATUS AND METHOD FOR CONTROLLING PRINTING APPARATUS - A printing apparatus that performs printing on a medium while being manually moved relative to the medium, the printing apparatus including a first discharger including a first nozzle row that discharges a first liquid, a second discharger including a second nozzle row that discharges a second liquid and is so provided as to be separate from the first nozzle row in a first direction perpendicular to the first nozzle row, a movement detection section that detects the moving direction of the printing apparatus viewed from the side facing the printing apparatus while the printing apparatus is moved, and an error process section that carries out an error process when printing using both the first and second dischargers is performed and the moving direction detected by the movement detection section deviates from the first direction. | 2021-11-18 |
| 20210357715 | SIGNATURE-BASED UNIQUE IDENTIFIER - The technology described herein generates a unique identifier for a visual media that comprises pre-printed visual indications on the visual media and a user's handwritten signature. The location of the signature on the visual media can be determined by including preprinted fiducial marks on the visual media. The fiducial markers act as landmarks that allow the size and location of the signature to be determined in absolute terms. The unique identifier is then stored in computer memory on a user-experience server. The user-experience server can associate the unique identifier with a digital asset, such as an image or video, designated by the user. When the unique identifier is provided to the user-experience server a second time, the digital asset can be retrieved and output to the computing device that provided the unique identifier. | 2021-11-18 |
| 20210357716 | Power Management - A device comprising: an antenna; a power harvesting circuit for harvesting power from a radio frequency field received at the antenna in order to power functions of the device; a communication unit coupled to the antenna for transmitting and receiving signals by means of the antenna, the communication device being configured to communicate according to a protocol in which a party to a communication session deems the session to have timed out if during a predetermined period it does not receive a signal from another party to the session; and a module comprising a processing circuit; the device being configured to interrupt the operation of the module when the communication unit is transmitting a signal by means of the antenna. | 2021-11-18 |
| 20210357717 | PASSIVE WIRELESS MAGNETIC FIELD CHARACTERISTIC SENSING TAG AND SYSTEM, AND MAGNETIC FIELD QUANTITY ACQUISITION METHOD - Provided is a passive wireless magnetic field characteristic sensing tag and system, and a magnetic field quantity acquisition method, which integrates a passive magnetic field quantity sensor in a passive electronic tag and cooperates with a reader and a host computer to build a magnetic field sensing system, so as to realize high-precision and high-intelligence monitoring of the magnetic field quantity. | 2021-11-18 |
| 20210357718 | IC CARD WITH FINGERPRINT RECOGNITION FUNCTION AND WORKING METHOD THEREOF - A working method for an IC card having a fingerprint recognition function, comprising: an IC card receiving and determining an instruction type from a terminal, and when determined that the received instruction is an application selection instruction, the IC card selecting an application and returning a response to the terminal; when determined that the received instruction is a processing option acquisition instruction, the IC card acquiring a user fingerprint information verification state according to the content of the instruction, and if verification is successful, returning to the terminal a processing option instruction response containing an application file locator list for which a personal identification number does not need to be verified; if verification fails, returning to the terminal a processing option instruction response containing an application file locator list for which a personal identification number must be verified; when determined that the received instruction is a record reading instruction, the IC card returning a record reading response to the terminal according to the record reading instruction, wherein the record reading response contains a method for verifying a card holder. Thus, the risk of a personal identification number being leaked is avoided, thus enhancing the security of a transaction, while also improving user experience. | 2021-11-18 |
| 20210357719 | REDOX ACTIVE POLYMER DEVICES AND METHODS OF USING AND MANUFACTURING THE SAME - The disclosed technology relates generally to apparatus comprising conductive polymers and more particularly to tag and tag devices comprising a redox-active polymer film, and method of using and manufacturing the same. In one aspect, an apparatus includes a substrate and a conductive structure formed on the substrate which includes a layer of redox-active polymer film having mobile ions and electrons. The conductive structure further includes a first terminal and a second terminal configured to receive an electrical signal therebetween, where the layer of redox-active polymer is configured to conduct an electrical current generated by the mobile ions and the electrons in response to the electrical signal. The apparatus additionally includes a detection circuit operatively coupled to the conductive structure and configured to detect the electrical current flowing through the conductive structure. | 2021-11-18 |
| 20210357720 | METHOD FOR PRODUCING HOSELINES AND PIPELINES WITH RFID CHIPS - A method for producing hoselines having at least the following working steps: a) providing a hoseline or pipeline blank ( | 2021-11-18 |
| 20210357721 | METAL FASTENER WITH EMBEDDED RFID TAG AND METHOD OF PRODUCTION - The present disclosure is generally directed to an RFID tag for use with a metal fastener where the fastener operates as the antenna of the RFID tag. The RFID tag includes a microchip for storing data. The chip is electrically coupled to the metal fastener in order to receive and transmit the RF signal, the metal fastener thereby operating as the antenna for the RFID tag. | 2021-11-18 |
| 20210357722 | ELECTRONIC DEVICE AND OPERATING METHOD FOR PERFORMING OPERATION BASED ON VIRTUAL SIMULATOR MODULE - Provided is a method, performed by an electronic device, of an operation based on a virtual simulator module, wherein the electronic device obtains a simulation parameter set for each of a plurality of operations for performing simulations with respect to the plurality of operations, obtains first performance information for each operation using a simulator module, wherein the first performance information indicates performance of an operation simulated based on the simulation parameter set, obtains second performance information for each operation based on the first performance information using a modeling module, wherein the second performance information indicates performance of the operation simulated in the simulator module, and performs an operation of the plurality of operations based on the first performance information and the second performance information. | 2021-11-18 |
| 20210357723 | Distributed Processing System and Distributed Processing Method - A distributed processing system includes a plurality of lower-order aggregation networks and a higher-order aggregation network. The lower-order aggregation networks include a plurality of distributed processing nodes disposed in a ring form. The distributed processing nodes generate distributed data for each weight of a neural network of an own node. The lower-order aggregation networks aggregate, for each lower-order aggregation network, the distributed data generated by the distributed processing nodes. The higher-order aggregation network generates aggregated data where the aggregation results of the lower-order aggregation networks are further aggregated, and distributes to the lower-order aggregation networks. The lower-order aggregation networks distribute the aggregated data distributed thereto to the distributed processing nodes belonging to the same lower-order aggregation network. The distributed processing nodes update weights of the neural network based on the distributed aggregated data. | 2021-11-18 |
| 20210357724 | SYSTEM AND METHOD FOR SIGNAL CONVERSION IN A NEURAL NETWORK - A system for signal conversion in a neural network can include: a processing module configured to control an output signal of the processing module to be a fixed value, when an input signal of the processing module is in a first interval; the processing module being configured to control the output signal to be in a preset non-linear relationship with the input signal, when the input signal is in a second interval; and where the output signal increases nonlinearly with the increase of the input signal and finally converges. | 2021-11-18 |
| 20210357725 | CORRELATIVE TIME CODING METHOD FOR SPIKING NEURAL NETWORKS - A computer-implemented method for classification of an input element to an output class in a spiking neural network may be provided. The method comprises receiving an input data set comprising a plurality of elements, identifying a set of features and corresponding feature values for each element of the input data set, and associating each feature to a subset of spiking neurons of a set of input spiking neurons of the spiking neural network. Furthermore, the method comprises also generating, by the input spiking neurons, spikes at pseudo-random time instants depending on a value of the feature for a given input element, and classifying an element into a class depending on a distance measure value between output spiking patterns at output spiking neurons of the spiking neural network and a predefined target pattern related to the class. | 2021-11-18 |
| 20210357726 | FUSION STRUCTURE AND METHOD OF CONVOLUTIONAL NEURAL NETWORK AND SPIKING NEURAL NETWORK - A fusion structure ( | 2021-11-18 |
| 20210357727 | ABNORMALITY DETECTION DEVICE - An abnormality detection device, method, or a storage medium acquires learning target data and monitoring target data, generates a state observer by using a variable in an input variable configuration, generates a threshold, calculates an abnormality degree by combining a second state observation value and the monitoring target data and inputting a combined result to the competitive neural network, and calculates a determination result. | 2021-11-18 |
| 20210357728 | SYNTHETIC DATA GENERATION APPARATUS BASED ON GENERATIVE ADVERSARIAL NETWORKS AND LEARNING METHOD THEREOF - A synthetic data generation apparatus according to an embodiment includes a generator for generating synthetic data from an input value, a first discriminator learned to distinguish between actual data and the synthetic data, a second discriminator learned to distinguish between the actual data and the synthetic data while satisfying differential privacy, and a third discriminator learned to distinguish between first synthetic data which is output from the generator learned by the first discriminator and second synthetic data which is output from the generator learned by the second discriminator. | 2021-11-18 |
| 20210357729 | SYSTEM AND METHOD FOR EXPLAINING THE BEHAVIOR OF NEURAL NETWORKS - A computing machine accesses a set of intermediate artificial neurons in a deep neural network. The deep neural network is fully or partially trained. The computing machine computes, for each artificial neuron in the set of intermediate artificial neurons, an influence score based on an average gradient of an output quantity of interest with respect to the artificial neuron across a plurality of inputs weighted by a probability of each input. The computing machine provides an output associated with the computed influence scores. | 2021-11-18 |
| 20210357730 | MULTI-SIZE CONVOLUTIONAL LAYER BACKGROUND - Systems and methods for improved convolutional layers for neural networks are disclosed. An improved convolutional layer can obtain at least two input feature maps of differing channel sizes. The improved convolutional layer can generate an output feature map for each one of the at least two input feature maps. Each input feature map can be applied to a convolutional sub-layer to generate an intermediate feature map. For each intermediate feature map, versions of the remaining intermediate feature maps can be resized to match the channel size of the intermediate feature map. For each intermediate feature map, an output feature map can be generated by combining the intermediate feature map and the corresponding resized versions of the remaining intermediate feature maps. | 2021-11-18 |
| 20210357731 | CONTROLLING AGENTS USING AMORTIZED Q LEARNING - Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment. One of the methods includes receiving a current observation; processing the current observation using a proposal neural network to generate a proposal output that defines a proposal probability distribution over a set of possible actions that can be performed by the agent to interact with the environment; sampling (i) one or more actions from the set of possible actions in accordance with the proposal probability distribution and (ii) one or more actions randomly from the set of possible actions; processing the current observation and each sampled action using a Q neural network to generate a Q value; and selecting an action using the Q values generated by the Q neural network. | 2021-11-18 |
| 20210357732 | NEURAL NETWORK ACCELERATOR HARDWARE-SPECIFIC DIVISION OF INFERENCE INTO GROUPS OF LAYERS - Neural network accelerator hardware-specific division of inference may be performed by operations including obtaining a computational graph and a hardware chip configuration. The operations also include dividing inference of the plurality of layers into a plurality of groups. Each group includes a number of sequential layers based on an estimate of duration and energy consumption by the hardware chip to perform inference of the neural network by performing the mathematical operations on activation data, sequentially by layer, of corresponding portions of layers of each group. The operations further include generating instructions for the hardware chip to perform inference of the neural network, sequentially by group, of the plurality of groups. | 2021-11-18 |
| 20210357733 | NEURAL NETWORK SECURITY - Herein is disclosed a neural network controller, configured to implement a neural network, the neural network including: a first layer; one or more second layers; and a third layer; wherein each layer of the first layer, the one or more second layers, and the third layer includes one or more nodes; wherein at least one node of the one or more second layers is configured to provide an output value at a first level of precision; wherein the neural network controller is configured to implement a precision reduction function to reduce an output value of at least one node of the third layer to a second level of precision; and wherein the second level of precision is less precise than the first level of precision. | 2021-11-18 |
| 20210357734 | Z-FIRST REFERENCE NEURAL PROCESSING UNIT FOR MAPPING WINOGRAD CONVOLUTION AND A METHOD THEREOF - A z-first reference neural processing unit (NPU) for mapping Winograd Convolution is disclosed where the NPU includes memory banks configured to store input feature maps (IFMs) in a z-first data storage layout, each of the memory banks being configured to store the IFMs in one of a direct convolution (DConv) mode or a Winograd convolution (WgConv) mode, a reconfigurable IFM distributor configured to receive the IFMs from the memory banks, a parallel reconfigurable Winograd forward transform module configured to receive the IFMs from the reconfigurable IFM distributor and to transform the IFMs in a Winograd domain to transformed IFMs in the WgConv mode, multiply and accumulate (MAC) units configured to perform dot product operations on one of IFMs in the DConv mode and the transformed IFMs in the WgConv mode to obtain intermediate output feature maps (OFMs), and a reconfigurable OFM adder and Winograd inverse transform module configured to generate one of an OFM from the intermediate OFMs in the DConv mode and OFMs from the intermediate OFMs in the WgConv. | 2021-11-18 |
| 20210357735 | SPLIT ACCUMULATOR FOR CONVOLUTIONAL NEURAL NETWORK ACCELERATOR - Disclosed embodiments relate to a split accumulator for a convolutional neural network accelerator, comprising: arranging original weights in a computation sequence and aligning by bit to obtain a weight matrix, removing slack bits in the weight matrix, allowing essential bits in each column of the weight matrix to fill the vacancies according to the computation sequence to obtain an intermediate matrix, removing null rows in the intermediate matrix, obtain a kneading matrix, wherein each row of the kneading matrix serves as a kneading weight; obtaining positional information of the activation corresponding to each bit of the kneading weight; divides the kneading weight by bit into multiple weight segments, processing summation of the weight segments and the corresponding activations according to the positional information, and sending a processing result to an adder tree to obtain an output feature map by means of executing shift-and-add on the processing result. | 2021-11-18 |
| 20210357736 | DEEP NEURAL NETWORK HARDWARE ACCELERATOR BASED ON POWER EXPONENTIAL QUANTIZATION - A deep neural network hardware accelerator comprises: an AXI-4 bus interface, an input cache area, an output cache area, a weighting cache area, a weighting index cache area, an encoding module, a configurable state controller module, and a PE array. The input cache area and the output cache area are designed as a line cache structure; an encoder encodes weightings according to an ordered quantization set, the quantization set storing the possible value of the absolute value of all of the weightings after quantization. During the calculation of the accelerator, the PE unit reads data from the input cache area and the weighting index cache area to perform shift calculation, and sends the calculation result to the output cache area. The accelerator uses shift operations to replace floating point multiplication operations, reducing the requirements for computing resources, storage resources, and communication bandwidth, and increasing the calculation efficiency of the accelerator. | 2021-11-18 |
| 20210357737 | Large-Scale Artificial Neural-Network Accelerators Based on Coherent Detection and Optical Data Fan-Out - Deep learning performance is limited by computing power, which is in turn limited by energy consumption. Optics can make neural networks faster and more efficient, but current schemes suffer from limited connectivity and the large footprint of low-loss nanophotonic devices. Our optical neural network architecture addresses these problems using homodyne detection and optical data fan-out. It is scalable to large networks without sacrificing speed or consuming too much energy. It can perform inference and training and work with both fully connected and convolutional neural-network layers. In our architecture, each neural network layer operates on inputs and weights encoded onto optical pulse amplitudes. A homodyne detector computes the vector product of the inputs and weights. The nonlinear activation function is performed electronically on the output of this linear homodyne detection step. Optical modulators combine the outputs from the nonlinear activation function and encode them onto optical pulses input into the next layer. | 2021-11-18 |
| 20210357738 | OPTIMIZING CAPACITY AND LEARNING OF WEIGHTED REAL-VALUED LOGIC - Maximum expressivity can be received representing a ratio between maximum and minimum input weights to a neuron of a neural network implementing a weighted real-valued logic gate. Operator arity can be received associated with the neuron. Logical constraints associated with the weighted real-valued logic gate can be determined in terms of weights associated with inputs to the neuron, a threshold-of-truth, and a neuron threshold for activation. The threshold-of-truth can be determined as a parameter used in an activation function of the neuron, based on solving an activation optimization formulated based on the logical constraints, the activation optimization maximizing a product of expressivity representing a distribution width of input weights to the neuron and gradient quality for the neuron given the operator arity and the maximum expressivity. The neural network of logical neurons can be trained using the activation function at the neuron, the activation function using the determined threshold-of-truth. | 2021-11-18 |
| 20210357739 | MEMORY DEVICE TO TRAIN NEURAL NETWORKS - Methods, systems, and apparatuses related to training neural networks are described. For example, data management and training of one or more neural networks may be accomplished within a memory device, such as a dynamic random-access memory (DRAM) device. Neural networks may thus be trained in the absence of specialized circuitry and/or in the absence of vast computing resources. One or more neural networks may be written or stored within memory banks of a memory device and operations may be performed within or adjacent to those memory banks to train different neural networks that are located in different banks of the memory device. This data management and training may occur within a memory system without involving a host device, processor, or accelerator that is external to the memory system. A trained network may then be read from the memory system and used for inference or other operations on an external device. | 2021-11-18 |
| 20210357740 | SECOND-ORDER OPTIMIZATION METHODS FOR AVOIDING SADDLE POINTS DURING THE TRAINING OF DEEP NEURAL NETWORKS - A computer-implemented method for training a deep neural network includes defining a loss function corresponding to the deep neural network, receiving a training dataset comprising training samples, and setting current parameter values to initial parameter values. An optimization method is performed which iteratively minimizes the loss function. During each iteration, a steepest direction of the loss function is calculated by determining the gradient of the loss function at the current parameter values. A batch of samples included in training samples is selected. A matrix-free CG solver is applied to obtain an inexact solution to a linear system defined by the steepest direction of the loss function and a stochastic Hessian matrix with respect to the batch of samples. A descent direction is determined, and the parameter values are updated based on the descent direction. Following the optimization method, the parameter values are stored in relationship to the deep neural network. | 2021-11-18 |
| 20210357741 | SYSTEM AND METHOD FOR ENERGY EFFICIENT SENSORS WITH COMPRESSION, ARTIFICIAL INTELLIGENCE, AND SECURITY - In a system and method for processing detected signals at a detector using a processor, a set of data is converted into a compressed set of data using a compressive sensing component controlled via a processor, the compressed set of data is transformed into a vector and the vector is filtered using a machine learning component controlled via the processor, the filtered vector is encrypted using an encryption component controlled via the processor, and the filtered vector is integrity protected using an integrity protection component controlled via the processor. | 2021-11-18 |
| 20210357742 | Real-Time Cognitive Wireless Networking Through Deep Learning in Transmission and Reception Communication Paths - Apparatuses and methods for real-time spectrum-driven embedded wireless networking through deep learning are provided. Radio frequency, optical, or acoustic communication apparatus include a programmable logic system having a front-end configuration core, a learning core, and a learning actuation core. The learning core includes a deep learning neural network that receives and processes input in-phase/quadrature (I/Q) input samples through the neural network layers to extract RF, optical, or acoustic spectrum information. A processing system having a learning controller module controls operations of the learning core and the learning actuation core. The processing system and the programmable logic system are operable to configure one or more communication and networking parameters for transmission via the transceiver in response to extracted spectrum information. | 2021-11-18 |
| 20210357743 | VARIATIONAL GRADIENT FLOW - According to embodiments of the present disclosure, methods of and computer program products for operating a plurality of classifiers are provided. A plurality of input entities are read, each input entity having an associated target label. The input entities are provided to a first classifier, and a category of each input entity is obtained therefrom. A feature map is determined for each input entity. Each feature map is provided to each of a set of classifiers, and an assigned label is obtained for each feature map from each of the set of classifiers. Each classifier is associated with one of the categories. For each classifier, the assigned label for each feature map is compared to the target labels to determine a plurality of gradients. The plurality of gradients are masked according to each category, yielding a masked set of gradients for each category. Each classifier is trained according its associated masked gradients. | 2021-11-18 |
