Learning Systems Patents (Class 382/155)
  • Patent number: 11392122
    Abstract: The technology relates to assisting large self-driving vehicles, such as cargo vehicles, as they maneuver towards and/or park at a destination facility. This may include a given vehicle transitioning between different autonomous driving modes. Such a vehicles may be permitted to drive in a fully autonomous mode on certain roadways for the majority of a trip, but may need to change to a partially autonomous mode on other roadways or when entering or leaving a destination facility such as a warehouse, depot or service center. Large vehicles such as cargo truck may have limited room to maneuver in and park at the destination, which may also prevent operation in a fully autonomous mode. Here, information from the destination facility and/or a remote assistance service can be employed to aid in real-time semi-autonomous maneuvering.
    Type: Grant
    Filed: August 23, 2019
    Date of Patent: July 19, 2022
    Assignee: Waymo LLC
    Inventors: Vijaysai Patnaik, William Grossman
  • Patent number: 11385901
    Abstract: A system including: at least one processor; and at least one memory having stored thereon computer program code that, when executed by the at least one processor, controls the system to: receive a data model identification and a dataset; in response to determining that the data model does not contain a hierarchical structure, perform expectation propagation on the dataset to approximate the data model with a hierarchical structure; divide the dataset into a plurality of channels; for each of the plurality of channels: divide the data into a plurality of microbatches; process each microbatch of the plurality of microbatches through parallel iterators; and process the output of the parallel iterators through single-instruction multiple-data (SIMD) layers; and asynchronously merge results of the SIMD layers.
    Type: Grant
    Filed: May 13, 2020
    Date of Patent: July 12, 2022
    Inventors: Matthew van Adelsberg, Rohit Joshi, Siqi Wang
  • Patent number: 11373285
    Abstract: An image generation means 81 generates an image using a generator. A discrimination means 82 discriminates whether an object image includes a feature of a target image, using a discriminator. A first update means 83 updates the generator so as to minimize a first error representing a degree of divergence between a result of discriminating a generated image using the discriminator and a correct answer label associated with the generated image, the generated image being the image generated using the generator. A second update means 84 updates the discriminator so as to minimize a second error representing a degree of divergence between each of respective results of discriminating the generated image, a first actual image including the feature of the target image, and a second actual image not including the feature of the target image using the discriminator and a correct answer label associated with a corresponding image.
    Type: Grant
    Filed: March 22, 2018
    Date of Patent: June 28, 2022
    Inventors: Kyota Higa, Azusa Sawada
  • Patent number: 11354792
    Abstract: Technologies for image processing based on a creation workflow for creating a type of images are provided. Both multi-stage image generation as well as multi-stage image editing of an existing image are supported. To accomplish this, one system models the sequential creation stages of the creation workflow. In the backward direction, inference networks can backward transform an image into various intermediate stages. In the forward direction, generation networks can forward transform an earlier-stage image into a later-stage image based on stage-specific operations. Advantageously, this technical solution overcomes the limitations of the single-stage generation strategy with a multi-stage framework to model different types of variation at various creation stages. Resultantly, both novices and seasoned artists can use these technologies to efficiently perform complex artwork creation or editing tasks.
    Type: Grant
    Filed: February 7, 2020
    Date of Patent: June 7, 2022
    Assignee: Adobe Inc.
    Inventors: Matthew David Fisher, Hung-Yu Tseng, Yijun Li, Jingwan Lu
  • Patent number: 11341361
    Abstract: An analysis method executed by a computer includes acquiring a refine image that maximizes a score for inferring a correct label by an inferring process using a trained model, the refine image being generated from an input image used when an incorrect label is inferred; generating a map indicating a region of pixels having the same or similar level of attention degree related to inference in the inferring process, of a plurality of pixels in the generated refine image, based on a feature amount used in the inferring process; extracting an image corresponding to a pixel region whose level in the generated map is a predetermined level, from calculated images calculated based on the input image and the refine image; and generating an output image that specifies a portion related to an inference error in the inferring process, among the calculated images, based on image processing on the extracted image.
    Type: Grant
    Filed: October 6, 2020
    Date of Patent: May 24, 2022
    Inventors: Tomonori Kubota, Takanori Nakao, Yasuyuki Murata
  • Patent number: 11328184
    Abstract: Disclosed are an image classification and conversion method, apparatus, image processor and training method thereof, and medium. The image classification method includes receiving a first input image and a second input image; performing image encoding on the first input image by utilizing n stages of encoding units connected in cascades to produce a first output image, wherein n is an integer greater than 1, and wherein as for 1?i<n, the output of the i-th stage of encoding unit is an input of an (i+1)-th stage of encoding unit, wherein m is an integer greater than 1; outputting a first output image, the first output image comprising mn output sub-images, and each of the mn output sub-images is corresponding to an image category.
    Type: Grant
    Filed: October 31, 2018
    Date of Patent: May 10, 2022
    Inventors: Pablo Navarrete Michelini, Hanwen Liu
  • Patent number: 11321529
    Abstract: A date extractor disclosed herein allows extracting dates and date ranges from documents. An implementation of the date extractor is implemented using various computer process instructions including scanning a document to generate a plurality of tokens, assigning labels to token using named entity recognition machine to generate a named entity vector, extracting dates from the named entity vector by comparing each of the named entities of the named entity vector to predetermined patterns of dates to generate a date vector, generating a plurality of date pairs from the date vector, and extracting date-ranges by comparing the plurality of date pairs to predetermined patterns of date ranges.
    Type: Grant
    Filed: December 25, 2018
    Date of Patent: May 3, 2022
    Assignee: Microsoft Technology Licensing, LLC
    Inventors: Ying Wang, Min Li, Mengyan Lu
  • Patent number: 11308319
    Abstract: A technique making use of a few-shot model to determine graphical features present in an image based on a small set of examples with known graphical features. Where a support set including a number of images that each have a known combination of graphical features, the image recognition can identify unknown combinations of those graphical features in any number of query images. In an embodiment of the present disclosure examples of a filled-out form are used to interpret any number of additional filled out versions of the form.
    Type: Grant
    Filed: November 27, 2019
    Date of Patent: April 19, 2022
    Assignee: DST Technologies, Inc.
    Inventors: Hui Peng Hu, Ramesh Sridharan
  • Patent number: 11301754
    Abstract: An information processing device and method for sharing of compressed training data for neural network training is provided. The information processing device receives a first image which includes an object of interest. The information processing device extracts, from the received first image, a region of interest which includes the object of interest. Once extracted, the extracted region of interest is provided to an input layer of N numbers of layers of a first neural network, trained on an object detection task. The information processing device selects an intermediate layer of the first neural network and extracts a first intermediate result as an output generated by the selected intermediate layer of the first neural network based on the input RoI. Once extracted, the information processing device shares the extracted first intermediate result as compressed training data with a server to train a second neural network on the object detection task.
    Type: Grant
    Filed: December 10, 2019
    Date of Patent: April 12, 2022
    Inventor: Nikolaos Georgis
  • Patent number: 11295172
    Abstract: Method of detecting objects in non-perspective images starts by generating an arrangement of tiles based on a field of view of a non-perspective camera lens, a predetermined size of the tiles, and a predetermined maximum object radius. The arrangement of the tiles includes the minimum number of tiles to cover the field of view. A non-perspective image is then captured using the non-perspective camera lens. The non-perspective image may be a still image frame or a video. Using the tiles, a plurality of images are generated, respectively, and at least a portion of a first object is detected in one or more images. The first object is generated using the one or more images that include the at least the portion of the first object, and the first object is displayed on a display interface. Other embodiments are described herein.
    Type: Grant
    Filed: May 21, 2019
    Date of Patent: April 5, 2022
    Assignee: Snap Inc.
    Inventors: Kevin Xie Chen, Shree K. Nayar
  • Patent number: 11288883
    Abstract: A method for controlling a robotic device is presented. The method includes capturing an image corresponding to a current view of the robotic device. The method also includes identifying a keyframe image comprising a first set of pixels matching a second set of pixels of the image. The method further includes performing, by the robotic device, a task corresponding to the keyframe image.
    Type: Grant
    Filed: September 13, 2019
    Date of Patent: March 29, 2022
    Inventors: Jeremy Ma, Kevin Stone, Max Bajracharya, Krishna Shankar
  • Patent number: 11263483
    Abstract: A method and apparatus for recognizing an image, and a storage medium are provided. The method includes: obtaining labeling results of training data to be labeled in a target training data set; determining a weight coefficient of each labeling result in response to that the training data includes a plurality of labeling results; determining a label of the training data based on labeling result and the weight coefficient corresponding to the labeling result, wherein the training data with the label is the target training data for training a recognition model; and recognizing the image based on the recognition model trained based on the target training data.
    Type: Grant
    Filed: May 29, 2020
    Date of Patent: March 1, 2022
    Assignee: Beijing Dajia Internet Information Technology Co., Ltd.
    Inventors: Zhiwei Zhang, Yan Li, Lijun Wu
  • Patent number: 11256899
    Abstract: A apparatus for classifying a seed as inbred or hybrid comprises a terahertz signal source for emitting a terahertz signal towards the seed, a detector for detecting at least part of the terahertz signal having interacted with the seed, a scanner for moving the support relative to the terahertz signal to provide a scan of the seed, a data processing device for forming an image data from the detected terahertz signal as obtained for a plurality of positions during the scan of the seed, and a decision support system for providing a classification from the image data. In an embodiment, the terahertz signal source is arranged for emitting a continuous or pulse wave signal, and wherein the detector is arranged for detecting an amplitude and a phase of the terahertz signal having interacted with the seed. A signal representing an outcome of the classification may control a separator to separate seeds according to their classification.
    Type: Grant
    Filed: January 15, 2015
    Date of Patent: February 22, 2022
    Inventors: Andrei Mikhailovitch Barychev, Alena Vladimirovna Belitskaya, Andrey Vyacheslavovich Khudchenko, Cornelia Catharina De Groot
  • Patent number: 11221484
    Abstract: A scent visualization system may comprise a display apparatus for generating a target image including a target object; an olfactory sensor for detecting a scent; and a scent visualization apparatus for generating target associative visualization information that reminds the scent from the target image received from the display apparatus and sensing information received from the olfactory sensor, and generating an associative image by combining the target associative visualization information and the target image. Therefore, a low-cost, high-efficiency, high-utilization, and high-convenience scent visualization system can be provided.
    Type: Grant
    Filed: October 14, 2019
    Date of Patent: January 11, 2022
    Assignee: Electronics and Telecommunications Research Institute
    Inventors: Jong Woo Choi, Jun Seok Park, Joon Hak Bang, Hae Ryong Lee, Sung June Chang
  • Patent number: 11210513
    Abstract: A computer-implemented detection method includes, in response to inputting a first image including a region of one or more objects to a learned model, identifying a first entire image corresponding to entirety of a first object as a detection candidate, the learned model being generated by learning training data including an image corresponding to a part of an object and an entire image corresponding to entirety of the object, detecting an existing region of the first target object in the first image in accordance with a comparison between the identified first entire image and the region of the one or more target objects, and determining, based on a specific image obtained by invalidating the existing region in the first image, whether another target object is included in the first image.
    Type: Grant
    Filed: October 22, 2019
    Date of Patent: December 28, 2021
    Inventors: Toshio Endoh, Keizo Kato
  • Patent number: 11176728
    Abstract: A method for rendering a non-photorealistic (NPR) content from a set (SI) of at least one image of a same scene is provided. The set of images (SI) is associated with a depth image comprising a set of regions. Each region corresponds to a region of a given depth. The method for rendering a non-photorealistic content includes generation of a segmented image having at least one segmented region generated with a given segmentation scale. The at least one segmented region corresponds to at least one region of the set of regions. A binary edge image is generated in which at least one binary edge region is generated with a given edge extraction scale, the at least one binary edge region corresponding to at least one region of the set of regions. The non-photorealistic content is rendered by combining the segmented image and the binary edge image.
    Type: Grant
    Filed: February 22, 2017
    Date of Patent: November 16, 2021
    Inventors: Caroline Baillard, Pierrick Jouet, Vincent Alleaume
  • Patent number: 11170020
    Abstract: Methods, computer systems, computer-storage media, and graphical user interfaces are provided for facilitating data transformations, according to embodiments of the present invention. In one embodiment, a plurality of remote sources is searched to identify candidate transformation tools relevant for performing data transformations. The candidate transformation tools are analyzed to identify tool examples corresponding with each of the candidate transformation tools. For each of the candidate transformation tools, the tool examples are stored in association with the corresponding candidate transformation tool. Based on a comparison of tool examples with example values, a transformation tool is identified as relevant to facilitate transforming example input values to the desired form in which to transform data.
    Type: Grant
    Filed: November 4, 2016
    Date of Patent: November 9, 2021
    Inventors: Yeye He, Kris Ganjam, Vivek Ravindranath Narasayya, Surajit Chaudhuri, Xu Chu
  • Patent number: 11163788
    Abstract: Methods, computer systems, computer-storage media, and graphical user interfaces are provided for facilitating data transformations, according to embodiments of the present invention. In one embodiment, a set of example values is received. An index to identify a plurality of data transformation tools that are relevant to the set of example values is referenced, wherein each of the data transformation tools correspond with one or more tool examples. The data transformation tools are ranked based on an extent of similarity between the set of example values and the tool examples. For data transformation tools associated with the extent of similarity that exceeds a similarity threshold, a transformation program is generated that uses the data transformation tool and a supplemental transformation tool to transform the one or more example input values to the desired form in which to transform data.
    Type: Grant
    Filed: November 4, 2016
    Date of Patent: November 2, 2021
    Inventors: Yeye He, Kris Ganjam, Vivek Ravindranath Narasayya, Surajit Chaudhuri, Xu Chu
  • Patent number: 11164438
    Abstract: Methods for detecting anomalies in a geographic area include receiving, from an electronic computing device, expected relationship data indicating expected relationships between a plurality of entities within the geographic area; detecting the plurality of entities within the geographic area; generating observed relationship data indicating observed relationships between the plurality of entities; identifying the expected relationships between the plurality of entities based on the expected relationship data; determining that a given observed relationship between two entities of the plurality of entities is likely to represent an anomaly based on the expected relationship data; and providing an electronic notification to a safety officer, the electronic notification indicating that the given observed relationship is likely to represent the anomaly.
    Type: Grant
    Filed: April 1, 2020
    Date of Patent: November 2, 2021
    Inventors: Dung-Han Lee, Yanyan Hu, Qifan Liang, Kangyan Liu, Yin Wang, Weijuan Wu, Chia Ying Lee
  • Patent number: 11151693
    Abstract: An image processing apparatus configured to divide an image into frequency bands and reduce noise, in which a first image includes a high-band frequency component, and a second image includes a low-band frequency component, includes: a first detecting unit configured to detect an edge in at least one of the first image and the second image; a second detecting unit configured to detect a low-contrast edge that is of a lower contrast than a contrast of the edge detected by the first detecting unit in the at least one of the first image and the second image; a compositing unit configured to composite the first image and the second image using a weighting corresponding to the edge and the low-contrast edge.
    Type: Grant
    Filed: July 30, 2018
    Date of Patent: October 19, 2021
    Inventor: Kazuhiro Yahata
  • Patent number: 11144832
    Abstract: A system and method for determining an optimal solution to an optimization problem in a swarm of candidate solutions is provided. The invention comprises generating a population of random particles, where each particle is representative of a candidate solution. Further, a best particle is identified from the generated population of particles. The best particle is representative of an optimal solution. The population of particles is categorised into similar and non-similar particle groups by applying one or more multivariate measurement techniques, and similarity between the particles of the non-similar particle group with best particle is updated by applying an imitation technique. The generated population is updated with updated particles and a new best particle is evaluated from said population. Furthermore, final best particle is determined by further updating the population of particles until one or more target conditions are achieved.
    Type: Grant
    Filed: January 23, 2018
    Date of Patent: October 12, 2021
    Inventors: Geelapaturu Subrahmanya Venkata Radha Krishna Rao, Gangadharan Ramakrishnan Gangadharan, Dinesh Reddy Vemula
  • Patent number: 11120072
    Abstract: A computer transforms high-dimensional data into low-dimensional data. (A) A distance matrix is computed from observation vectors. (B) A kernel matrix is computed from the distance matrix using a bandwidth value. (C) The kernel matrix is decomposed using an eigen decomposition to define eigenvalues. (D) A predefined number of largest eigenvalues are selected from the eigenvalues. (E) The selected largest eigenvalues are summed. (F) A next bandwidth value is computed based on the summed eigenvalues. (A) through (F) are repeated with the next bandwidth value until a stop criterion is satisfied. Each observation vector of the observation vectors is transformed into a second space using a kernel principal component analysis with the next bandwidth value and the kernel matrix. The second space has a dimension defined by the predefined number of first eigenvalues. Each transformed observation vector is output.
    Type: Grant
    Filed: February 23, 2021
    Date of Patent: September 14, 2021
    Assignee: SAS Institute Inc.
    Inventors: Kai Shen, Haoyu Wang, Arin Chaudhuri
  • Patent number: 11109085
    Abstract: The present disclosure relates to training a recommendation model to generate trait recommendations using one permutation hashing and populated-value-slot-based densification. In particular, the disclosed systems can train the recommendation model by computing sketch vectors corresponding to traits using one permutation hashing. The disclosed systems can then fill in unpopulated value slots of the sketch vectors using populated-value-slot-based densification. The disclosed systems can combine the resulting densified sketches to generate the trained recommendation model. For example, in some embodiments, the disclosed systems can combine the sketches by generating a plurality of locality sensitive hashing tables based on the sketches. In some embodiments, the disclosed systems generate a count sketch matrix based on the sketches and generate trait embeddings based on the count sketch matrix using spectral embedding.
    Type: Grant
    Filed: March 28, 2019
    Date of Patent: August 31, 2021
    Assignee: ADOBE INC.
    Inventors: Anup Rao, Yasin Abbasi Yadkori, Tung Mai, Ryan Rossi, Ritwik Sinha, Matvey Kapilevich, Alexandru Ionut Hodorogea
  • Patent number: 11100366
    Abstract: Methods and systems for digital image segmentation and annotation, including: receiving a digital image depicting, in part, an object of interest from an input file; one or more of manually and automatically adding a polygon around the object of interest to generate a segmented digital image; one or more of manually and automatically appending a label to the polygon around the object of interest to generate a segmented and annotated digital image, wherein the label indicates one or more of an identity and a characteristic of the object of interest; and outputting information related to the segmented and annotated digital image to an output file. Optionally, the polygon is one of a bounding box and a 4-point polygon. Optionally, the object of interest is a parking spot.
    Type: Grant
    Filed: April 26, 2018
    Date of Patent: August 24, 2021
    Assignee: Volvo Car Corporation
    Inventors: Sohini Roy Chowdhury, Andreas Wallin, Sihao Ding
  • Patent number: 11093690
    Abstract: A computing system accesses an image-based document and a text document having text extracted from the image-based document and provides a user interface displaying at least a portion of the image-based document. In response to selection of a text portion of the image-based document, the system determines an occurrence of the text portion within at least a portion of the image-based document and then applies a search model on the text document to identify the same occurrence of the text portion. Once matched, alignment data indicating a relationship between a selected tag and both the text portion of the image-based document and the text portion of the text document is stored.
    Type: Grant
    Filed: August 14, 2019
    Date of Patent: August 17, 2021
    Assignee: Palantir Technologies Inc.
    Inventors: Suchan Lee, Jon Paek
  • Patent number: 11087156
    Abstract: A system and method for detecting, identifying, and displaying handwriting-based entry is provided. The system and method include features for detecting entry of at least one first letter based on handwriting, identifying a style of the at least one first letter, and displaying at least one second letter associated with the at least one first letter based on, and in the form of, the identified style of the at least one first letter.
    Type: Grant
    Filed: June 14, 2019
    Date of Patent: August 10, 2021
    Assignee: Samsung Electronics Co., Ltd.
    Inventors: Olha Zubarieva, Tetiana Ignatova, Ivan Deriuga, Olga Radyvonenko, Serhii Polotskyi, Vadym Osadchiy, Oleksandr Shchur
  • Patent number: 11055580
    Abstract: The disclosure relates to a method and an apparatus for analyzing an image using a deep neural net pre-trained for multiple classes. The image is processed by a forward pass through an adapted neural net to generate a processing result. The adapted neural net is adapted from the pre-trained neural net to focus on exactly one selected class. The processing result is then analyzed focused on features corresponding to the selected class using an image processing algorithm. A modified image is generated by removing a manifestation of these features from the image.
    Type: Grant
    Filed: June 4, 2018
    Date of Patent: July 6, 2021
    Inventors: Peter Amon, Andreas Hutter, Sanjukta Ghosh
  • Patent number: 11055821
    Abstract: The presently-disclosed technology improves the resolution of an x-ray microscope so as to obtain super-resolution x-ray images having resolutions beyond the maximum normal resolution of the x-ray microscope. Furthermore, the disclosed technology provides for the rapid generation of the super-resolution x-ray images and so enables real-time super-resolution x-ray imaging for purposes of defect detection, for example. A method of super-resolution x-ray imaging using a super-resolving patch classifier is provided. In addition, a method of training the super-resolving patch classifier is disclosed. Other embodiments, aspects and features are also disclosed.
    Type: Grant
    Filed: May 20, 2020
    Date of Patent: July 6, 2021
    Assignee: SVXR, Inc.
    Inventors: Edward R. Ratner, David L. Adler
  • Patent number: 11049007
    Abstract: A recognition apparatus based on a deep neural network, a training apparatus and methods thereof. The deep neural network is obtained by inputting training samples comprising positive samples and negative samples into an input layer of the deep neural network and training. The apparatus includes: a judging unit configured to judge that a sample to be recognized is a suspected abnormal sample when confidences of positive sample classes in a classification result outputted by an output layer of the deep neural network are all less than a predefined threshold value. Hence, reliability of a confidence of a classification result outputted by the deep neural network may be efficiently improved.
    Type: Grant
    Filed: May 5, 2017
    Date of Patent: June 29, 2021
    Inventors: Song Wang, Wei Fan, Jun Sun
  • Patent number: 11037024
    Abstract: In an aspect of the present disclosure relates to a network involving humans and AI based systems working in conjunction to perform tasks such as traffic violation detection, infrastructure monitoring, traffic flow management, crop monitoring etc. from visual data acquired from numerous data acquisition sources. The system includes a network of electronic mobile devices with AI capabilities, connected to a decentralized network working towards capturing high quality data for finding events or objects of interest in the real world, retraining AI models, for processing the high volumes of data on decentralized or centralized processing units and also being used for the verification of the outputs of the AI models. The system talks about many annotation techniques on smartphones for crowdsourced AI Data Labeling for AI Training.
    Type: Grant
    Filed: November 29, 2019
    Date of Patent: June 15, 2021
    Inventor: Jayant Ratti
  • Patent number: 11036703
    Abstract: There is disclosed a method and system to operate a software application entirely based on a unitary lexicon data structure (LDS) comprising a plurality of data field definition blocks stored in memory, with one LDS record for each lexicon term. The LDS is used to develop computerized lexicons and deploy them for use to operate a lexical application with all data displayed for viewing and input by the user on a single screen to which all desired data items come, rather than the user navigating to fields statically located on a multitude of screens. Each LDS record contains a whole set of data in memory, with data duplicated across LDS records in order to bypass the need for the application to interoperate with a database to input and display related data. There is a graphical icon also of a unitary format into which all data is input and displayed. Input data items are related to one another in the icon, regardless of whether a relational database is configured to interoperate with the system.
    Type: Grant
    Filed: August 7, 2018
    Date of Patent: June 15, 2021
    Inventor: Thomas Graham Shaughnessy
  • Patent number: 11017255
    Abstract: The present disclosure relates to an image analysis method, system, and computer program. The image analysis method of the present disclosure includes: receiving a query image; extracting one or more regions of interest from the query image; calculating a first feature for each of the regions of interest by respectively applying the regions of interest to one or more ROI (region of interest) feature extraction models independently learned in order to extract features of the regions of interest; and calculating analysis values of the query image by applying the first features of the regions of interest to a pre-learned integration analysis model. According to the present disclosure, it is possible to reduce the influence on an analysis model by an error that training data created for map learning of an entire image may have, and it is also possible to increase learning accuracy and objectivity of a deep neural network.
    Type: Grant
    Filed: August 9, 2018
    Date of Patent: May 25, 2021
    Inventor: Jae Joon Lee
  • Patent number: 10977528
    Abstract: Techniques for determining image similar are described. For example, a computer-implemented method comprising: receiving a request to determine similarity between a first image and at least one other image; determining similarity between the first image and at least one other image based upon one or more Gram matrix-based style values and one or more vector distance calculation-based content values as determined from one or more outputs of layers of a convolutional neural network; and providing an indication of the similarity of between the first image and the at least one other image is described.
    Type: Grant
    Filed: June 25, 2019
    Date of Patent: April 13, 2021
    Assignee: Amazon Technologies, Inc.
    Inventor: Dylan Tong
  • Patent number: 10977740
    Abstract: Disclosed is a system and method to automatically identify property-related risks through the use of computer vision, sensors, and/or building information models (BIMs). The ability to automatically identify a variety of hazards helps mitigate the associated risks, and thus reduces the number of accidents or fatalities that would otherwise occur. In some embodiments, a “risk map” can be generated by mapping the identified risks for a given property.
    Type: Grant
    Filed: April 30, 2020
    Date of Patent: April 13, 2021
    Inventors: Siddhartha Dalal, Devasis Bassu, Promiti Dutta, Ashwath Aithal, Liangkai Zhang, Alvin Chou, Michael P. Castelli
  • Patent number: 10970530
    Abstract: Techniques for grammar-based automated generation of annotated synthetic form training data for machine learning are described. A training data generation engine utilizes a defined grammar to construct a layout for a form, select key-value units to place within the layout, and select attribute variants for the key-value units. The form is rendered and stored at a storage location, where it can be provided along with other similarly-generated forms to be used as training data for a machine learning model.
    Type: Grant
    Filed: November 13, 2018
    Date of Patent: April 6, 2021
    Assignee: Amazon Technologies, Inc.
    Inventors: Amit Adam, Oron Anschel, Or Perel, Gal Sabina Star, Omri Ben-Eliezer, Hadar Averbuch Elor, Shai Mazor, Wendy Tse, Andrea Olgiati, Rahul Bhotika, Stefano Soatto
  • Patent number: 10963690
    Abstract: A method and device for identifying a main picture in a web page. The method comprises: picking out candidate main pictures based on a page attributes of each picture in a web page (210); cropping an original picture of each candidate main picture to obtain a corresponding picture composition (220); determining a candidate main picture having an information topic matching a topic of the web page (230); and identifying a picture composition corresponding to the matched candidate main picture as the main picture of the web page (240).
    Type: Grant
    Filed: May 9, 2017
    Date of Patent: March 30, 2021
    Inventors: Shouke Qin, You Han, Peizhi Xu, Xuezhong Qiu, Xiaolin Ma
  • Patent number: 10963741
    Abstract: The invention relates to a control device (1) for a vehicle for determining the perceptual load of a visual and dynamic driving scene. The control device is configured to: receive a sensor output (101) of a sensor (3), the sensor (3) sensing the visual driving scene, extract a set of scene features (102) from the sensor output (101), the set of scene features (102) representing static and/or dynamic information of the visual driving scene, and determine the perceptual load (104) of the set of extracted scene features (102) based on a predetermined load model (103), wherein the load model (103) is predetermined based on reference video scenes each being labelled with a load value The invention further relates to a system and a method.
    Type: Grant
    Filed: June 7, 2016
    Date of Patent: March 30, 2021
    Inventors: Jonas Ambeck-Madsen, Ichiro Sakata, Nilli Lavie, Gabriel J. Brostow, Luke Palmer, Alina Bialkowski
  • Patent number: 10958828
    Abstract: A method for improving the performance of a computer vision system includes obtaining input specifying a task to be performed by the computer vision system; obtaining a first digital image; and comparing the first digital image to at least one training image used to train the computer vision system to solve the task. Further steps include, based on the comparing indicating that the first digital image is insufficiently similar to the at least one training image, recommending at least one adjustment to the digital image; obtaining a second digital image in accordance with the adjustment; and performing the task with the computer vision system based on the second digital image obtained in accordance with the adjustment. Adjustments can be based, for example, on image composition and/or weather conditions.
    Type: Grant
    Filed: October 10, 2018
    Date of Patent: March 23, 2021
    Assignee: International Business Machines Corporation
    Inventors: Marco Aurelio Stelmar Netto, Andrea Britto Mattos Lima, Maysa Malfiza Garcia de Macedo, Maciel Zortea, Igor Cerqueira Oliveira
  • Patent number: 10924800
    Abstract: Disclosed are systems and methods for improving interactions with and between computers in content generating, searching, hosting and/or providing systems supported by or configured with personal computing devices, servers and/or platforms. The systems interact to identify and retrieve data within or across platforms, which can be used to improve the quality of data used in processing interactions between or among processors in such systems. The disclosed systems and methods provide systems and methods for automatically detecting and rendering highlights from streaming videos in real-time. As a streaming video is being broadcast over the Internet, the disclosed systems and methods determine each type of scene from the streaming video, and automatically score highlight scenes. The scored highlight scenes are then communicated to users as compiled video segments, which can be over any type of channel or platform accessible to a user's device and network that enables content rendering and user interaction.
    Type: Grant
    Filed: August 19, 2019
    Date of Patent: February 16, 2021
    Assignee: VERIZON MEDIA INC.
    Inventors: Yale Song, Jordi Vallmitjana
  • Patent number: 10915799
    Abstract: An image processing apparatus includes a memory and a processor coupled to the memory. The processor is configured to classify each of a plurality of images into one of a plurality of groups based on a feature of each of the plurality of images. The processor is configured to store first information in the memory. A first recognition method is associated with a first group in the first information and images classified into the first group are correctly recognized by the first recognition method. The processor is configured to store second information in the memory. One of at least one second recognition method different from the first recognition method is associated with a second group in the second information and images classified into the second group are incorrectly recognized by the first recognition method.
    Type: Grant
    Filed: March 11, 2019
    Date of Patent: February 9, 2021
    Inventors: Yuji Matsuda, Eigo Segawa
  • Patent number: 10896355
    Abstract: Disclosed are systems and methods for automatic selection of canonical digital images from a large corpus of digital images, such as the corpus of digital images available on the web, for an entity, such as and without limitation a person, a point of interest, object, etc. The automated, unsupervised approach for selecting a diverse set of high quality, canonical digital images, is well suited for processing a large corpus of digital images. A set of canonical digital images identified for an entity can be retrieved in response to a digital image request for digital images depicting the entity.
    Type: Grant
    Filed: December 21, 2018
    Date of Patent: January 19, 2021
    Assignee: VERIZON MEDIA INC.
    Inventors: Sachin Sudhakar Farfade, Vijay Mahadevan, Ayman Kaheel, Ayyappan Arasu, Venkat Kumar Reddy Barakam, Jay Kiran Mahadeokar
  • Patent number: 10891488
    Abstract: Described is a system for visual activity recognition. In operation, the system detects a set of objects of interest (OI) in video data and determines an object classification for each object in the set of OI, the set including at least one OI. A corresponding activity track is formed for each object in the set of OI by tracking each object across frames. Using a feature extractor, the system determines a corresponding feature in the video data for each OI, which is then used to determine a corresponding initial activity classification for each OI. One or more OI are then detected in each activity track via foveation, with the initial object detection and foveated object detection thereafter being appended into a new detected-objects list. Finally, a final classification is provided for each activity track using the new detected-objects list and filtering the initial activity classification results using contextual logic.
    Type: Grant
    Filed: January 14, 2019
    Date of Patent: January 12, 2021
    Assignee: HRL Laboratories, LLC
    Inventors: Deepak Khosla, Ryan M. Uhlenbrock, Huapeng Su, Yang Chen
  • Patent number: 10885464
    Abstract: Relevance decay techniques are provided for time-based evaluation of machine learning applications and other classifiers. An exemplary method comprises obtaining time series measurement data; generating an input dataset comprising a plurality of records, wherein each record comprises features extracted from the time series measurement data, a target class corresponding to an event to be identified, and a time lag indicating a difference in time between a given extraction and the event to be identified; evaluating a plurality of classifiers during an evaluation phase using a portion of the input dataset and one or more predefined evaluation metrics weighted using a time-based relevance decay function based on the time lag; and selecting one or more of the classifiers to perform classification of the time series measurement data based on the predefined weighted evaluation metrics during a classification phase.
    Type: Grant
    Filed: October 31, 2016
    Date of Patent: January 5, 2021
    Assignee: EMC IP Holding Company LLC
    Inventors: Diego Salomone Bruno, Victor Bursztyn, Percy E. Rivera Salas, Tiago Salviano Calmon
  • Patent number: 10878601
    Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a final classification output for an image of eye tissue. The image is provided as input to each of one or more segmentation neural networks to obtain one or more segmentation maps of the eye tissue in the image. A respective classification input is generated from each of the segmentation maps. For each of the segmentation maps, the classification input for the segmentation map is provided as input to each of one or more classification neural networks to obtain, for each segmentation map, a respective classification output from each classification neural network. A final classification output for the image is generated from the respective classification outputs for each of the segmentation maps.
    Type: Grant
    Filed: December 28, 2018
    Date of Patent: December 29, 2020
    Assignee: Google LLC
    Inventors: Jeffrey De Fauw, Joseph R. Ledsam, Bernardino Romera-Paredes, Stanislav Nikolov, Nenad Tomasev, Samuel Blackwell, Harry Askham, Xavier Glorot, Balaji Lakshminarayanan, Trevor Back, Mustafa Suleyman, Pearse A. Keane, Olaf Ronneberger, Julien Robert Michel Cornebise
  • Patent number: 10872230
    Abstract: Methods and systems may provide for facial recognition of at least one input image utilizing hierarchical feature learning and pair-wise classification. Receptive field theory may be used on the input image to generate a pre-processed multi-channel image. Channels in the pre-processed image may be activated based on the amount of feature rich details within the channels. Similarly, local patches may be activated based on the discriminant features within the local patches. Features may be extracted from the local patches and the most discriminant features may be selected in order to perform feature matching on pair sets. The system may utilize patch feature pooling, pair-wise matching, and large-scale training in order to quickly and accurately perform facial recognition at a low cost for both system memory and computation.
    Type: Grant
    Filed: March 27, 2015
    Date of Patent: December 22, 2020
    Assignee: Intel Corporation
    Inventors: Jianguo Li, Yurong Chen, Ke Chen, Yi-Jen Chiu
  • Patent number: 10861198
    Abstract: A method of assisting in selection of a particular mask from among a plurality of masks, the method comprising: generating a plurality of augmented faces, each augmented face corresponding to a respective mask of the plurality of masks; and associating each augmented face with a respective mask of the plurality of masks.
    Type: Grant
    Filed: June 29, 2018
    Date of Patent: December 8, 2020
    Assignee: Koninklijke Philips N.V.
    Inventor: Dmitry Nikolayevich Znamenskiy
  • Patent number: 10853691
    Abstract: A method includes processing an input image using convolution layers to define image features and processing the image features to define feature statistics. Processing the image features includes applying an activation function in a feature dimension of the image features. The method also includes processing the feature statistics using fully connected layers to produce a binary output regarding a characteristic of the input image.
    Type: Grant
    Filed: January 7, 2019
    Date of Patent: December 1, 2020
    Assignee: Apple Inc.
    Inventors: Russell Y. Webb, Jason Ramapuram
  • Patent number: 10827185
    Abstract: In various embodiments, a quality trainer trains a model that computes a value for a perceptual video quality metric for encoded video content. During a pre-training phase, the quality trainer partitions baseline values for metrics that describe baseline encoded video content into partitions based on genre. The quality trainer then performs cross-validation operations on the partitions to optimize hyperparameters associated with the model. Subsequently, during a training phase, the quality trainer performs training operations on the model that includes the optimized hyperparameters based on the baseline values for the metrics to generate a trained model. The trained model accurately tracks the video quality for the baseline encoded video content. Further, because the cross-validation operations minimize any potential overfitting, the trained model accurately and consistently predicts perceived video quality for non-baseline encoded video content across a wide range of genres.
    Type: Grant
    Filed: July 11, 2016
    Date of Patent: November 3, 2020
    Assignee: NETFLIX, INC.
    Inventors: Anne Aaron, Zhi Li, Todd Goodall
  • Patent number: 10809080
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining routing. An exemplary method comprises: inputting a plurality of to-be-optimized routing solution candidates to a Siamese neural network comprising a plurality of value prediction networks, each of the value prediction networks being trained to predict a cost associated with a to-be-optimized routing solution candidate; identifying one or more to-be-optimized routing solution candidates from the plurality of to-be-optimized routing solution candidates based on outputs of the Siamese neural network; inputting the one or more identified to-be-optimized routing solution candidates to a routing optimizer to obtain one or more optimized routing solution candidates; and determining an optimized routing solution with a lowest cost from the one or more optimized routing solution candidates.
    Type: Grant
    Filed: March 23, 2020
    Date of Patent: October 20, 2020
    Inventors: Xingwen Zhang, Shuang Yang
  • Patent number: 10776926
    Abstract: A system and method for training a computer-implemented object classifier includes detecting a foreground visual object within a sub-region of a scene, determining a background model of the sub-region of the scene, the background model representing the sub-region when any foreground visual object is absent from that sub-region, and training the object classifier by computer-implemented machine learning using the background model of the sub-region as a negative training example.
    Type: Grant
    Filed: March 14, 2017
    Date of Patent: September 15, 2020
    Assignee: Avigilon Corporation
    Inventor: Ashish Shrivastava