Trainable Classifiers Or Pattern Recognizers (e.g., Adaline, Perceptron) Patents (Class 382/159)
  • Patent number: 10121076
    Abstract: An entity interaction recognition system algorithmically recognizes a variety of different types of entity interactions that may be captured in two-dimensional images. In some embodiments, the system estimates the three-dimensional spatial configuration or arrangement of entities depicted in the image. In some embodiments, the system applies a proxemics-based analysis to determine an interaction type. In some embodiments, the system infers, from a characteristic of an entity detected in an image, an area or entity of interest in the image.
    Type: Grant
    Filed: May 2, 2016
    Date of Patent: November 6, 2018
    Assignee: SRI International
    Inventors: Ishani Chakraborty, Hui Cheng, Omar Javed
  • Patent number: 10121094
    Abstract: A system, method and computer program product is provided. An input signal for classification and a set of pre-classified signals are received, each comprising a vector representation of an object having a plurality of vector elements. A sparse vector comprising a plurality of sparse vector coefficients is determined. Each sparse vector coefficient corresponds to a signal in the set of pre-classified signals and represents the likelihood of a match between the object represented in the input signal and the object represented in the corresponding signal. A largest sparse vector coefficient is compared with a predetermined threshold. If the largest sparse vector coefficient is less than the predetermined threshold, the corresponding signal is removed from the set of pre-classified signals. The determining and comparing are repeated using the input signal and the reduced set of pre-classified signals.
    Type: Grant
    Filed: December 9, 2016
    Date of Patent: November 6, 2018
    Assignee: International Business Machines Corporation
    Inventors: Cecilia J. Aas, Raymond S. Glover
  • Patent number: 10121055
    Abstract: This invention describes methods and systems for the automated facial landmark localization. Our approach proceeds from sparse to dense landmarking steps using a set of models to best account for the shape and texture variation manifested by facial landmarks across pose and expression. We also describe the use of an l1-regularized least squares approach that we incorporate into our shape model, which is an improvement over the shape model used by several prior Active Shape Model (ASM) based facial landmark localization algorithms.
    Type: Grant
    Filed: September 8, 2016
    Date of Patent: November 6, 2018
    Assignee: CARNEGIE MELLON UNIVERSITY
    Inventors: Marios Savvides, Keshav Thirumalai Seshadri
  • Patent number: 10111632
    Abstract: For breast cancer detection with an x-ray scanner, a cascade of multiple classifiers is trained or used. One or more of the classifiers uses a deep-learnt network trained on non-x-ray data, at least initially, to extract features. Alternatively or additionally, one or more of the classifiers is trained using classification of patches rather than pixels and/or classification with regression to create additional cancer-positive partial samples.
    Type: Grant
    Filed: January 31, 2017
    Date of Patent: October 30, 2018
    Assignee: Siemens Healthcare GmbH
    Inventors: Yaron Anavi, Atilla Peter Kiraly, David Liu, Shaohua Kevin Zhou, Zhoubing Xu, Dorin Comaniciu
  • Patent number: 10108848
    Abstract: This invention relates to a method of analyzing a factor of an attribute based on a case sample set containing combinations of image data and attribute data associated with the image data. The attribute factor analysis method includes: a division step of dividing an image region of the image data forming each element of the case sample set into parts in a mesh shape of a predetermined sample size; a reconstruction step of reconstructing, based on the case sample set, the case sample sets for the respective parts to obtain reconstructed case sample sets; an analysis step of analyzing, for each of the reconstructed case sample sets, a dependency between an explanatory variable representing a feature value of image data on each part and an objective variable representing the attribute data, to thereby obtain an attribute factor analysis result; and a visualization step of visualizing the attribute factor analysis result to produce the visualized attribute factor analysis result.
    Type: Grant
    Filed: March 28, 2014
    Date of Patent: October 23, 2018
    Assignee: NEC SOLUTION INNOVATORS, LTD.
    Inventors: Yasuyuki Ihara, Masashi Sugiyama
  • Patent number: 10102256
    Abstract: A method and system for improving an Internet based search is provided. The method includes generating an intent domain associated with a subject based intent classification. An unstructured data analysis process is executed with respect to a content corpus being associated with the subject based intent classification and a search phase entered in a search field of a graphical user interface with respect to a domain specific search query for specified subject matter. In response the subject based intent classification is determined to be associated with the search query and the subject based intent classification is compared to search results data. A subset of search results of the search results data correlating to the subject based intent classification is determined and ranked resulting in a ranked list. The subject based intent classification and the ranked list are presented to a user.
    Type: Grant
    Filed: September 29, 2016
    Date of Patent: October 16, 2018
    Assignee: International Business Machines Corporation
    Inventors: Gilbert Barron, Jasmine S. Basrai, Michael J. Bordash, Lisa Seacat DeLuca
  • Patent number: 10104302
    Abstract: An image determining method, applied to an image sensing apparatus with a lens and an image sensor comprising a first pixel and a second pixel. First/second parts of the first/second pixels are respectively covered. The image determining method comprises: applying the first/second pixel to respectively generate a first/second object image, wherein the first/second object images are combined to generate a first combined object image; computing a first brightness information variation tendency of the first object image, a second brightness information variation tendency of the second object image and brightness information variation tendency of the first combined object image; and determining if the first combined object image is a front image or a rear image based on the first brightness information variation tendency, the second brightness information variation tendency and the brightness information variation tendency of the first combined object image.
    Type: Grant
    Filed: May 22, 2016
    Date of Patent: October 16, 2018
    Assignee: PixArt Imaging Inc.
    Inventor: Guo-Zhen Wang
  • Patent number: 10083343
    Abstract: At least some example embodiments disclose a method and apparatus for facial recognition. The facial recognition method includes detecting initial landmarks from a facial image, first normalizing the facial image using the initial landmarks, updating a position of at least one of intermediate landmarks based on the first normalizing, the intermediate landmarks being landmarks transformed from the initial landmarks through the first normalizing, second normalizing the facial image after the updating and recognizing a face using a feature of the second normalized facial image.
    Type: Grant
    Filed: July 9, 2015
    Date of Patent: September 25, 2018
    Assignee: SAMSUNG ELECTRONICS CO., LTD.
    Inventors: Sungjoo Suh, Wonjun Hwang, Wonjun Kim, Jaejoon Han
  • Patent number: 10067659
    Abstract: A method and system for identifying a rotational orientation of an implanted electrical stimulation lead utilize radiological images of the lead. The lead has an asymmetric marker with a longitudinal band extending around a portion of the circumference of the lead. The method and system includes obtaining radiological images of the lead; generating an isosurface image from the radiological images and displaying the isosurface image on a display device, where the isosurface image comprises an image of the longitudinal band of the marker; identifying a bulge in the isosurface image corresponding the longitudinal band of the marker; and determining a rotational orientation of the lead based on the rotational orientation of the bulge in the isosurface image.
    Type: Grant
    Filed: August 23, 2016
    Date of Patent: September 4, 2018
    Assignee: Boston Scientific Neuromodulation Corporation
    Inventor: Hemant Bokil
  • Patent number: 10067760
    Abstract: A system and method for classifying and resolving software production incident tickets includes receiving an incident ticket, extracting a plurality of keywords from the incident ticket, and deriving a query vector corresponding to the incident ticket based on the plurality of keywords. The system and method further comprises classifying the incident ticket into at least one of a positive mechanization incident ticket and a negative mechanization incident ticket based on a comparison of the query vector and a plurality of vectors derived from a plurality of past incident tickets. The plurality of vectors are derived based on a plurality of keywords and their corresponding occurrences in the plurality of past incident tickets.
    Type: Grant
    Filed: March 14, 2016
    Date of Patent: September 4, 2018
    Assignee: Wipro Limited
    Inventors: Premchand Ryali, Shivamurthy Harave Guruswamappa, Ramkumar Balasubramanian
  • Patent number: 10068321
    Abstract: When a human skin region is detected, misdetection or detection failure is prevented. An image processing apparatus of the present invention acquires an image from which a human body skin region is to be extracted, and estimates a human body part region in the image. The image processing apparatus then estimates a skin attribute of the estimated human body part region and corrects the human body part region based on the estimated skin attribute. The image processing apparatus extracts the skin region from the acquired image based on the skin attribute of the estimated human body part region and the corrected human body part region.
    Type: Grant
    Filed: November 19, 2015
    Date of Patent: September 4, 2018
    Assignee: Canon Kabushiki Kaisha
    Inventors: Koichi Magai, Masakazu Matsugu, Shunta Tate
  • Patent number: 10068154
    Abstract: Recognition process (1) of an object (An) in a query image (2), performing a training step (6) that comprises: —providing (20) a set of training images (Ti), each training image (Ti) comprising an object tag (LOGOi); —determining (21) for each training image (Ti) of said set a plurality of first descriptors (11), each first descriptor (11) being a vector that represents pixel properties in a subregion (Sri) of the associated training image (Ti); —determining (23) a group of exemplar descriptors (111) describing the set of training images (Ti) and resulting from a selection of the first descriptors (11) based on the position of said subregion (Sri) in the associated training image (Ti) and on the pixel properties of said first descriptors (11); performing a query step (7) comprising: —receiving (30) the query image (2) and defining (31) a plurality of vectors (V?) of second descriptors (3) describing the properties of said query image (2); —determining (35) a visual similarity coefficient based on a comparison
    Type: Grant
    Filed: August 13, 2013
    Date of Patent: September 4, 2018
    Assignee: LOGOGRAB LIMITED
    Inventor: Alessandro Prest
  • Patent number: 10062151
    Abstract: An image deblurring method and an image deblurring apparatus are provided. The image deblurring method includes acquiring a blurred image and a dynamic vision sensor (DVS) event set that is recorded by a DVS while the blurred image is exposed, and deblurring the blurred image, based on the DVS event set.
    Type: Grant
    Filed: November 21, 2016
    Date of Patent: August 28, 2018
    Assignee: SAMSUNG ELECTRONICS CO., LTD.
    Inventors: Qiang Wang, Kyoobin Lee, Keun Joo Park, Ping Guo
  • Patent number: 10062007
    Abstract: An apparatus stores a plurality of partial programs, which are constituent elements of an image recognizing program that detects a position of a template image on an input image. The apparatus creates a plurality of individual programs each being a combination of at least two of the plurality of partial programs, and calculates a similarity map that associates similarity with the template image with each pixel of the input image by using each of the plurality of individual programs. The apparatus calculates fitness for each of the plurality of individual programs, based on a distribution of the similarity map, selects an individual program for which the fitness is equal to or greater than a prescribed threshold, from among the plurality of individual programs, and outputs the selected individual program as the image recognizing program.
    Type: Grant
    Filed: June 2, 2016
    Date of Patent: August 28, 2018
    Assignee: FUJITSU LIMITED
    Inventors: Tsuyoshi Nagato, Tetsuo Koezuka
  • Patent number: 10055666
    Abstract: A real-time, semi-automatic method for discriminant track initialization in thermal imagery is very critical for tracking since it determines what to track for the tracker. Therefore, any insignificant or false information may result in redundant features or deceptive appearance that can cause premature track losses. To achieve long-term tracking, a real-time, semi-automatic track initialization methodology for IR images is proposed which takes a single image coordinate as input, then generates target bounding box for the most salient segment. The present invention is designed for real-time applications in which erroneous user input is unavoidable. Therefore, error compensation capability of present invention both introduces many practical advantageous together with performance boost. Low computational cost and requirement of only a point coordinate as input in the vicinity of the target make this approach preferable in real-time tracking applications.
    Type: Grant
    Filed: May 8, 2014
    Date of Patent: August 21, 2018
    Assignee: ASELSAN ELEKTRONIK SANAYI VE TICARET ANONIM SIRKETI
    Inventors: Emre Tunali, Sinan Oz
  • Patent number: 10055801
    Abstract: A system and method for efficiently and securely converting physical documents, including handwritten reports, into structured data stored in databases is disclosed. The method includes extraction by a processor of electronic images or text from scanned physical documents and transmission of the extracted data by the processor to other devices for evaluation by data entry technicians. A candidate set of potential responses to a field may be provided to a data entry technician to enable the data entry technician to evaluate the extracted data. The candidate set may be generated based on responses extracted physical documents having the same author. Entries from different data fields of documents may be provided to different data entry technicians to reduce the amount of information provided to a data entry technician, thereby reducing security risks for documents containing confidential material.
    Type: Grant
    Filed: June 23, 2016
    Date of Patent: August 21, 2018
    Assignee: LIBERTY PIPELINE SERVICES, LLC
    Inventors: Charles Eric Smith, Benjamin Dominguez-Benner
  • Patent number: 10049306
    Abstract: Aspects of the present disclosure involve a system and method for learning from images of transactional data. In one embodiment, a system is introduced that can learn from the images of transactional data. In particular, machine learning is implemented on images in order to classify information in a more accurate manner. The images are created from raw data deriving from a user account.
    Type: Grant
    Filed: December 29, 2016
    Date of Patent: August 14, 2018
    Assignee: PAYPAL, INC.
    Inventors: Lian Liu, Hui-Min Chen
  • Patent number: 10037128
    Abstract: A method and apparatus for generating, causing display of, sending, or otherwise providing one or more interfaces for monitoring machines, classifying operating behavior of the machines, and/or predicting operational behavior of the machines. The interfaces may include a graphical user interface that that shows graphical elements changing in appearance over time according to a sequence of stored values that represent measurements. An interface manager may receive, via the graphical user interface, input comprising a label, and, based at least in part on the input, store the label in association with the displayed sequence or in association with a set or cluster of sequences that includes the displayed sequence.
    Type: Grant
    Filed: February 4, 2014
    Date of Patent: July 31, 2018
    Assignee: Falkonry, Inc.
    Inventors: Nikunj R. Mehta, Prasanta Bose
  • Patent number: 10038710
    Abstract: A first set of log entries is identified. A plurality of log entry classes occurring in the first set of log entries is determined. Each log entry in a given log entry class has a same number, type, and ordering of components. A vector of component type identifiers is determined for each log entry class. Each identifier in a vector for a given log entry class identifies a position and type of a component included in a log entry belonging to the given log entry class. A classification tree is created using the vectors. An unclassified log entry not included in the first set of log entries is identified. A log entry class is assigned to the unclassified log entry using the classification tree to create a classified log entry. One or more security threat patterns are evaluated using the classified log entry.
    Type: Grant
    Filed: December 22, 2015
    Date of Patent: July 31, 2018
    Assignee: SAP SE
    Inventor: Viktor Bersch
  • Patent number: 10033979
    Abstract: A video surveillance system, device and methods may accurately model the shape of a human object monitored by a video stream. 3D human models, such as a coarse 3D human model and a detailed 3D human model may be estimated by mapping individual body part components to a frame. For example, a coarse 3D human model may be obtained by mapping the cylindrical body parts to a plurality of skeleton pose estimates on a part by part basis. A detailed 3D human model may be estimated by mapping detailed human body parts to respective the cylindrical body parts of the coarse 3D human model on a part by part basis. The detailed 3D human model may be used to detect accessories of the human object being monitored, as well as overall dimensions, body part dimensions, age, and gender of the human object being monitored.
    Type: Grant
    Filed: March 15, 2013
    Date of Patent: July 24, 2018
    Assignee: AVIGILON FORTRESS CORPORATION
    Inventors: Atul Kanaujia, Niels Haering, Mun Wai Lee
  • Patent number: 10032482
    Abstract: A CPU extracts still images from image files recorded at predetermined time intervals among still images and moving images arbitrarily recorded in a predetermined period such as one day, and generates a time-lapse moving image by taking these extracted images as frames. That is, in addition to still images, even when a moving image file is recorded at timing corresponding to a predetermined time interval, a frame at the timing corresponding to the predetermined time interval in the moving image file is extracted, and adopted as a frame of a time-lapse moving image to be generated.
    Type: Grant
    Filed: March 23, 2016
    Date of Patent: July 24, 2018
    Assignee: Casio Computer Co., Ltd.
    Inventor: Kenji Yoshizawa
  • Patent number: 10026017
    Abstract: A method for generating labels for an image of low quality is described. The method includes mapping image data and depth information of said image to a 3D point cloud; segmenting the 3-D point cloud and the image into super voxels and image patches, fusing features obtained from the super voxels and image patches, by using a fusion model, applying classifiers to fused features, wherein the fusion model and the classifiers are generated from a dataset including image data of selected quality and quantity, corresponding point cloud and image labels, and generating scene labels based on applied classifiers.
    Type: Grant
    Filed: October 10, 2016
    Date of Patent: July 17, 2018
    Assignee: THOMSON LICENSING
    Inventor: Tao Luo
  • Patent number: 10019655
    Abstract: Systems and methods are disclosed herein for automatically identifying a query object within a visual medium. The technique generally involves receiving as input to a neural network a query object and a visual medium including the query object. The technique also involves generating, by the neural network, representations of the query object and the visual medium defining features of the query object and the visual medium. The technique also involves generating, by the neural network, a heat map using the representations. The heat map identifies a location of pixels corresponding to the query object within the visual medium and is usable to generate an updated visual medium highlighting the query object.
    Type: Grant
    Filed: August 31, 2016
    Date of Patent: July 10, 2018
    Assignee: Adobe Systems Incorporated
    Inventors: Balaji Krishnamurthy, Mausoom Sarkar
  • Patent number: 10019535
    Abstract: The disclosed embodiments provide a system that processes data. During operation, the system obtains text from a document associated with a user. Next, the system applies a set of rules to each word in the text to determine a context associated with the word. The system then extracts data associated with the context from the text. Finally, the system enables use of the data with one or more applications without requiring manual input of the data into the one or more applications.
    Type: Grant
    Filed: August 6, 2013
    Date of Patent: July 10, 2018
    Assignee: INTUIT INC.
    Inventors: Sunil H. Madhani, Anu Sreepathy, Samir Revti Kakkar
  • Patent number: 10013727
    Abstract: A report concerning the contents obtained by interpretation based on a medical image can be efficiently created without any constraints of expression. An information processing apparatus according to this invention includes an image analysis unit which acquires information concerning a region name or disease name based on an analysis result on the input medical image, an input unit which inputs the result obtained by interpreting the medical image as character information, a conversion candidate prediction unit which outputs conversion candidates concerning the input character information, and a display control unit which displays the input character information upon converting the character information into character information selected from the conversion candidates. The apparatus further includes a priority level setting unit which sets priority levels in advance for character information output as the conversion candidates.
    Type: Grant
    Filed: August 28, 2008
    Date of Patent: July 3, 2018
    Assignee: CANON KABUSHIKI KAISHA
    Inventors: Yoshihiko Iwase, Akihiro Katayama, Hiroshi Imamura
  • Patent number: 10013639
    Abstract: Digital images, such as digital photographs, are analyzed by an application running on a user device or other computing apparatus. Heuristics, characteristic detection or measurement techniques, or other analytics are used to evaluate individual digital images or to compare a plurality of digital images in accordance with user-input criteria. Digital images are then presented to a user as a result of the analysis, and further operations may be performed per user selections or input. Numerous digital images may thus be timely evaluated for aesthetic appeal, composition, subject matter content, or other factors, and then deleted, printed, distributed, or put other use.
    Type: Grant
    Filed: December 16, 2013
    Date of Patent: July 3, 2018
    Assignee: AMAZON TECHNOLOGIES, INC.
    Inventors: Eric Michael Schurman, William Alexander Strand, Jo-Pete Nelson, Yaodong Liu
  • Patent number: 10007659
    Abstract: A method includes acquiring a first corpus, including first text of a first sentence including a first word and described in a natural language, and second text of a second sentence including a second word different in meaning from the first word, a second word distribution of the second word being similar to a first word distribution of the first word, acquiring a second corpus including third text of a third sentence, including a third word identical to the first word and/or the second word, a third word distribution of the third word being not similar to the first word distribution, based on an arrangement of a word string in the first corpus and the second corpus, assigning to the first word a first vector representing a meaning of the first word and assigning to the second word a second vector representing a meaning of the second word.
    Type: Grant
    Filed: June 7, 2016
    Date of Patent: June 26, 2018
    Assignee: PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO., LTD.
    Inventors: Katsuyoshi Yamagami, Takashi Ushio, Yasunori Ishii
  • Patent number: 10003895
    Abstract: Disclosed herein are methods, systems, and devices for selecting a scene classification for the operation of a sensory prosthesis, such as a hearing prosthesis. A system of two or more sensory prostheses can receive respective inputs from the environment of a recipient. A scene classification can then be determined for each sensory prosthesis based on the audio input received by each hearing prosthesis. A confidence value can also be determined for each scene classification. A scene classification can then be selected for each sensory prosthesis, from the determined scene classifications, based on the determined confidence values. Such operation can allow each sensory prosthesis to operate according to a respective selected scene classification that could be the same or that could be different from scene classifications selected for other sensory prostheses of the system.
    Type: Grant
    Filed: May 26, 2016
    Date of Patent: June 19, 2018
    Assignee: Cisco Technology, Inc.
    Inventors: Stephen Fung, Alexander von Brasch, Michael Goorevich
  • Patent number: 9996768
    Abstract: Neural network patch aggregation and statistical techniques are described. In one or more implementations, patches are generated from an image, e.g., randomly, and used to train a neural network. An aggregation of outputs of patches processed by the neural network may be used to label an image using an image descriptor, such as to label aesthetics of the image, classify the image, and so on. In another example, the patches may be used by the neural network to calculate statistics describing the patches, such as to describe statistics such as minimum, maximum, median, and average of activations of image characteristics of the individual patches. These statistics may also be used to support a variety of functionality, such as to label the image as described above.
    Type: Grant
    Filed: November 19, 2014
    Date of Patent: June 12, 2018
    Assignee: ADOBE SYSTEMS INCORPORATED
    Inventors: Xiaohui Shen, Xin Lu, Zhe Lin, Radomir Mech
  • Patent number: 9990539
    Abstract: An image recognition method and device are disclosed. The disclosed device may include: a reference image storage unit configured to convert reference images into 1-dimensional vectors and store the converted 1-dimensional vectors; an image acquisition unit configured to obtain an input image; an image converter unit configured to convert the input image into a 1-dimensional vector, apply dimension reduction, and calculate coefficients for expressing the input image as a linear sum of vectors forming a reference image for each of the reference images; and a determiner unit configured to determine a reference image corresponding to the input image by performing a subtraction computation of subtracting a value obtained by applying the coefficients calculated for each reference image to a relevant reference image from the converted input image. The invention makes it possible to perform image recognition at a fast speed and lower the likelihood of the users' privacy being invaded.
    Type: Grant
    Filed: November 2, 2015
    Date of Patent: June 5, 2018
    Assignee: INDUSTRY-ACADEMIC FOUNDATION, YONSEI UNIVERSITY
    Inventors: Min Ho Ka, Ji Woong Yu
  • Patent number: 9983553
    Abstract: An object of the present invention is to provide an autonomous system that realizes expected operation in a form in which the soundness of the operation can be proved to a third party in adaptation to an external factor that dynamically varies and to enhance a working ratio in autonomous operation. The autonomous system is provided with a function for dynamically leading a satisfiable combination of a requirement for the soundness of operation and expected operation on the basis of the information of operating environment acquired via exterior world measurement means, a function for generating control logic for realizing the expected operation, a function for recording the control logic, the requirement for sound operation and the expected operation, and a function for presenting the record in a form in which the third party can read the record.
    Type: Grant
    Filed: March 11, 2013
    Date of Patent: May 29, 2018
    Assignee: Hitachi, Ltd.
    Inventor: Masataka Nishi
  • Patent number: 9978002
    Abstract: System and method for determining a classifier to discriminate between two classes—object or non-object. The classifier may be used by an object detection program to detect presence of a 3D object in a 2D image. The overall classifier is constructed of a sequence of classifiers, where each such classifier is based on a ratio of two graphical probability models. A discreet-valued variable representation at each node in a Bayesian network by a two-stage process of tree-structured vector quantization is discussed. The overall classifier may be part of an object detector program that is trained to automatically detect different types of 3D objects. Computationally efficient statistical methods to evaluate overall classifiers are disclosed. The Bayesian network-based classifier may also be used to determine if two observations belong to the same category.
    Type: Grant
    Filed: November 10, 2015
    Date of Patent: May 22, 2018
    Assignee: Carnegie Mellon University
    Inventor: Henry Schneiderman
  • Patent number: 9964499
    Abstract: An apparatus for processing multi-energy image data to separate at least two types of material comprises a classification unit, wherein the classification unit is configured to obtain a classification of pixels or voxels belonging to the types of material based on a threshold which is adaptively changed in dependence on multi-energy intensity information associated with the pixels or voxels.
    Type: Grant
    Filed: November 4, 2014
    Date of Patent: May 8, 2018
    Assignee: Toshiba Medical Systems Corporation
    Inventors: Saad Masood, Costas Plakas
  • Patent number: 9965705
    Abstract: Described herein are systems and methods for generating and using attention-based deep learning architectures for visual question answering task (VQA) to automatically generate answers for image-related (still or video images) questions. To generate the correct answers, it is important for a model's attention to focus on the relevant regions of an image according to the question because different questions may ask about the attributes of different image regions. In embodiments, such question-guided attention is learned with a configurable convolutional neural network (ABC-CNN). Embodiments of the ABC-CNN models determine the attention maps by convolving image feature map with the configurable convolutional kernels determined by the questions semantics. In embodiments, the question-guided attention maps focus on the question-related regions and filters out noise in the unrelated regions.
    Type: Grant
    Filed: June 16, 2016
    Date of Patent: May 8, 2018
    Assignee: Baidu USA LLC
    Inventors: Kan Chen, Jiang Wang, Wei Xu
  • Patent number: 9953393
    Abstract: An analyzing method and an analyzing system for graphics process are provided. The analyzing method includes the following steps. A graphics application program is provided and a plurality of graphics parameters of the graphics application program are obtained. The graphics application program is classified to be at least one of a plurality of groups according to the graphics parameters. A plurality weighting coefficients are obtained. A total loading of a graphics processing unit for performing the graphics application program is calculated according to the weighting coefficients and the graphics parameters.
    Type: Grant
    Filed: December 29, 2015
    Date of Patent: April 24, 2018
    Assignee: INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE
    Inventors: Arthur Marmin, Chun-Hung Lai, Hsun-Lun Huang, Juin-Ming Lu
  • Patent number: 9948887
    Abstract: Exemplary embodiments for visual communication between a first device and a second device, comprising: creating a 3D mesh model of a first device user; receiving sensor data from a sensor array during the visual communication session between the first device and the second device, wherein the image data includes motion of the first device user; determining 3D mesh model updates using the image data; transmitting the 3D mesh model updates to the second device for the second device to update display of the 3D mesh model of the first device user, wherein the update is represented as one or more of a blend shape and a relative vertex position change of the 3D mesh model.
    Type: Grant
    Filed: March 17, 2016
    Date of Patent: April 17, 2018
    Assignee: Samsung Electronics Co., Ltd.
    Inventor: Schuyler Cullen
  • Patent number: 9940323
    Abstract: A mechanism is provided in a data processing system for text classification. A domain terms selector component, executing on a processor of the data processing system, receives an input text. A domain terms selector component executing on a processor of the data processing system identifies at least one domain term in the input text. A domain terms replacer component executing on a processor of the data processing system replaces the at least one identified domain term in the input text with a corresponding replacement term to form a revised input text. A text classifier component configured with a trained text classifier machine learning model classifies the revised input text to form a class determination.
    Type: Grant
    Filed: July 12, 2016
    Date of Patent: April 10, 2018
    Assignee: International Business Machines Corporation
    Inventors: John M. Boyer, Kshitij P. Fadnis, Dinesh Raghu
  • Patent number: 9934434
    Abstract: An image forensics system estimates a camera response function (CRF) associated with a digital image, and compares the estimated CRF to a set of rules and compares the estimated CRF to a known CRF. The known CRF is associated with a make and a model of an image sensing device. The system applies a fusion analysis to results obtained from comparing the estimated CRF to a set of rules and from comparing the estimated CRF to the known CRF, and assesses the integrity of the digital image as a function of the fusion analysis.
    Type: Grant
    Filed: June 30, 2016
    Date of Patent: April 3, 2018
    Assignee: Honeywell International Inc.
    Inventor: Scott McCloskey
  • Patent number: 9928442
    Abstract: Systems, methods, and computer program products to perform an operation comprising assigning each of a plurality of images in a blog post and each of a plurality of images in a collection of images to a respective node in a graph, computing an adjacency matrix for the graph, wherein the adjacency matrix defines relationships between images in the blog post and images in the collections of images, and determining a first subset of the images in the collection of images that summarize the images in the image collection, wherein the subset is determined based on the adjacency matrix, wherein the adjacency matrix is computed based on the subset of the images in the collection of images.
    Type: Grant
    Filed: March 22, 2016
    Date of Patent: March 27, 2018
    Assignee: Disney Enterprises, Inc.
    Inventors: Leonid Sigal, Gunhee Kim, Seungwhan Moon
  • Patent number: 9928447
    Abstract: The examiner has taken the exact language in WO 2015/094370 A1 to transcribe it: Systems, apparatus, and methods to determine relationships, group memberships, and social networks and circles automatically through an analysis of media content, such as photos or videos, and ascertaining the connections between individuals or groups. Embodiments include identifying individuals in photos with or without other people in different situations, and developing a context based on objects in the photo. The extracted context may be stored in a predefined ontology structure in a database and analyzed to determine relationships or group memberships. A database of automatically determined relationships may be analyzed in order to provide relationship recommendations, social advice, or to support law enforcement investigations. The database may be continuously updated with additional information provided by individuals or obtained through a secondary social network that includes user media or social information.
    Type: Grant
    Filed: December 20, 2013
    Date of Patent: March 27, 2018
    Assignee: Intel Corporation
    Inventors: Oded Koren, Shahar Taite, Yaniv Golan, Tomer Rider
  • Patent number: 9916524
    Abstract: Techniques for determining depth for a visual content item using machine-learning classifiers include obtaining a visual content item of a reference light pattern projected onto an object, and determining shifts in locations of pixels relative to other pixels representing the reference light pattern. Disparity, and thus depth, for pixels may be determined by executing one or more classifiers trained to identify disparity for pixels based on the shifts in locations of the pixels relative to other pixels of a visual content item depicting in the reference light pattern. Disparity for pixels may be determined using a visual content item of a reference light pattern projected onto an object without having to match pixels between two visual content items, such as a reference light pattern and a captured visual content item.
    Type: Grant
    Filed: March 15, 2016
    Date of Patent: March 13, 2018
    Assignee: Microsoft Technology Licensing, LLC
    Inventors: Sean Ryan Francesco Fanello, Christoph Rhemann, Adarsh Prakash Murthy Kowdle, Vladimir Tankovich, David Kim, Shahram Izadi
  • Patent number: 9905054
    Abstract: Techniques for controlling patch-usage in image synthesis are described. In implementations, a curve is fitted to a set of sorted matching errors that correspond to potential source-to-target patch assignments between a source image and a target image. Then, an error budget is determined using the curve. In an example, the error budget is usable to identify feasible patch assignments from the potential source-to-target patch assignments. Using the error budget along with uniform patch-usage enforcement, source patches from the source image are assigned to target patches in the target image. Then, at least one of the assigned source patches is assigned to an additional target patch based on the error budget. Subsequently, an image is synthesized based on the source patches assigned to the target patches.
    Type: Grant
    Filed: June 9, 2016
    Date of Patent: February 27, 2018
    Inventors: Jakub Fiser, Ondrej Jamri{hacek over (s)}ka, Michal Luká{hacek over (c)}, Elya Shechtman, Paul J. Asente, Jingwan Lu, Daniel Sýkora
  • Patent number: 9892431
    Abstract: A real-time messaging platform allows advertiser accounts to pay to insert candidate messages into the message streams requested by account holders. To accommodate multiple advertisers, the messaging platform controls an auction process that determines which candidate messages are selected for inclusion in a requested account holder's message stream. Selection is based on a bid for the candidate message, the message stream that is requested, and a variety of other factors that vary depending upon the implementation. The process for selection of candidate messages generally includes the following steps, though any given step may be omitted or combined into another step in a different implementation: targeting, filtering, prediction, ranking, and selection.
    Type: Grant
    Filed: September 8, 2016
    Date of Patent: February 13, 2018
    Assignee: Twitter, Inc.
    Inventors: Parag Agrawal, Utkarsh Srivastava, Kumar Chellapilla, Qicheng Ma
  • Patent number: 9875397
    Abstract: At least one example embodiment discloses a method of extracting a feature of an input image. The method includes constructing an example pyramid including at least one hierarchical level based on stored example images, generating a codebook in each of the at least one hierarchical level, calculating a similarity between the codebook and the input image, and extracting a feature of the input image based on the similarity.
    Type: Grant
    Filed: August 4, 2015
    Date of Patent: January 23, 2018
    Assignee: SAMSUNG ELECTRONICS CO., LTD.
    Inventors: Wonjun Hwang, Wonjun Kim, Sungjoo Suh, Jungbae Kim, Jaejoon Han
  • Patent number: 9858472
    Abstract: The present disclosure provides a three-dimensional facial recognition method and system. The method includes: performing pose estimation on an input binocular vision image pair by using a three-dimensional facial reference model, to obtain a pose parameter and a virtual image pair of the three-dimensional facial reference model with respect to the binocular vision image pair; reconstructing a facial depth image of the binocular vision image pair by using the virtual image pair as prior information; detecting, according to the pose parameter, a local grid scale-invariant feature descriptor corresponding to an interest point in the facial depth image; and generating a recognition result of the binocular vision image pair according to the detected local grid scale-invariant feature descriptor and training data having attached category annotations. The present disclosure can reduce computational costs and required storage space.
    Type: Grant
    Filed: July 18, 2016
    Date of Patent: January 2, 2018
    Assignee: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED
    Inventors: Yue Ming, Jie Jiang, Tingting Liu, Juhong Wang
  • Patent number: 9852340
    Abstract: A method of identifying, with a camera, an object in an image of a scene, by determining the distinctiveness of each of a number of attributes of an object of interest, independent of the camera viewpoint, determining the detectability of each of the attributes based on the relative orientation of a candidate object in the image of the scene, determining a camera setting for viewing the candidate object based on the distinctiveness of an attribute, so as to increase the detectability of the attribute, and capturing an image of the candidate object with the camera setting to determine the confidence that the candidate object is the object of interest.
    Type: Grant
    Filed: September 29, 2015
    Date of Patent: December 26, 2017
    Assignee: Canon Kabushiki Kaisha
    Inventors: Fei Mai, Geoffrey Richard Taylor
  • Patent number: 9852363
    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating labeled images. One of the methods includes selecting a plurality of candidate videos from videos identified in a response to a search query derived from a label for an object category; selecting one or more initial frames from each of the candidate videos; detecting one or more initial images of objects in the object category in the initial frames; for each initial frame including an initial image of an object in the object category, tracking the object through surrounding frames to identify additional images of the object; and selecting one or more images from the one or more initial images and one or more additional images as database images of objects belonging to the object category.
    Type: Grant
    Filed: January 5, 2016
    Date of Patent: December 26, 2017
    Assignee: Google Inc.
    Inventors: Jonathon Shlens, Quoc V. Le, Gregory Sean Corrado, Marc'Aurelio Ranzato
  • Patent number: 9832359
    Abstract: A focusing method for an image capturing device includes determining a first focus distance according to a first focus frame of the plurality of focus frames; capturing a first image according to the first focus frame, determining a second focus distance according to a second focus frame of the plurality of focus frames, and detecting whether a second depth of field corresponding to the second focus distance is overlapping with a first depth of field corresponding to the first focus distance, for determining whether to capture a second image according to the second focus distance.
    Type: Grant
    Filed: January 8, 2014
    Date of Patent: November 28, 2017
    Assignee: NOVATEK Microelectronics Corp.
    Inventor: Wei Hsu
  • Patent number: 9830502
    Abstract: In a method and system for identifying objects in an image, an image and training data are received. The training data identifies a pixel associated with an object of a particular type in the image. A plurality of filtered versions of the image are developed. The training data and the plurality of filtered versions of the image are processed to develop a trained model for classifying pixels associated with objects of the particular type. The trained model is applied to the image to identify pixels associated a plurality of objects of the particular type in the image. Additional image processing steps are developed to further refine the identified pixels for better fitting of the contour of the objects with their edges.
    Type: Grant
    Filed: October 14, 2014
    Date of Patent: November 28, 2017
    Assignee: DH Technologies Development Pte. Ltd.
    Inventors: Avrum Isaac Cohen, Dihui Hong
  • Patent number: 9830380
    Abstract: Embodiments provide a multimodality-based image tagging apparatus and a method for the same. The image tagging apparatus includes: a score generating unit configured to generate, for an inquiry image, multiple groups of first scores about all tags in an tagging dictionary by using a training image and multiple modalities of an image; a late-fusion unit configured to fuse the obtained multiple groups of scores to obtain final scores about all the tags; and a tag selecting unit configured to select one or more tag(s) with relatively large tag scores as tag(s) of the inquiry image according to the final scores about all the tags. With the embodiments, multiple modalities may be effectively fused, and a more robust and accurate image tagging result may be obtained.
    Type: Grant
    Filed: June 18, 2014
    Date of Patent: November 28, 2017
    Assignee: FUJITSU LIMITED
    Inventors: Xi Liu, Rujie Liu