Patents Examined by Xiao Liu
-
Patent number: 11328517Abstract: A system and method generate feature space data that may be used for object detection. The system includes one or more processors and a memory. The memory may include one or more modules having instructions that, when executed by the one or more processors, cause the one or more processors to obtain a two-dimension image of a scene, generate an output depth map based on the two-dimension image of the scene, generate a pseudo-LIDAR point cloud based on the output depth map, generate a bird's eye view (BEV) feature space based on the pseudo-LIDAR point cloud, and modify the BEV feature space to generate an improved BEV feature space using feature space neural network that was trained by using a training LIDAR feature space as a ground truth based on a LIDAR point cloud.Type: GrantFiled: May 20, 2020Date of Patent: May 10, 2022Assignee: Toyota Research Institute, Inc.Inventors: Victor Vaquero Gomez, Rares A. Ambrus, Vitor Guizilini, Adrien David Gaidon
-
Patent number: 11328522Abstract: In a learning device, method, and program for a discriminator, and a discriminator, it is possible to enable accurate learning of a discriminator that discriminates a state of an object to be observed, such as a cell. An image acquisition unit acquires a first image including an influence of a meniscus and a second image with the influence of the meniscus eliminated for the same object to be observed. Next, a training data generation unit generates training data for learning a discriminator based on the second image. Then, a learning unit learns the discriminator based on the first image and the training data.Type: GrantFiled: January 28, 2020Date of Patent: May 10, 2022Assignee: FUJIFILM CorporationInventor: Takayuki Tsujimoto
-
Patent number: 11294953Abstract: A similar face retrieval method and device and a storage medium are provided. The method includes: acquiring a face image to be retrieved; performing feature extraction on the face image to be retrieved based on a feature extraction model, and taking an output result of an antepenultimate layer of the feature extraction model as a first similar face feature of the face image to be retrieved; for each face image in a face database, acquiring the distance in feature space between the first similar face feature and a second similar face feature of each face image; and according to the obtained distance information, retrieving a target face image with the highest similarity to the face image to be retrieved in the face database.Type: GrantFiled: March 27, 2020Date of Patent: April 5, 2022Assignee: Beijing Dajia Internet Information Technology Co., Ltd.Inventors: Fan Yang, Yan Li
-
Patent number: 11288840Abstract: Disclosed is an artificial intelligence (AI) apparatus including a two-dimensional (2D) image sensor configured to acquire a 2D image of a head of a person, a three-dimensional (3D) image sensor configured to acquire 3D head pose information of the head, and a processor configured to match the 2D image with the 3D head pose information, to extract 3D head pose information for determining a rotation direction of the head from the 3D head pose information, to extract a 2D image matched with the extracted 3D head pose information, to acquire 3D relative coordinates as a reference for correcting the 3D head pose information based on 2D coordinates of a predetermined landmark point of the extracted 2D image, and to acquire the corrected 3D head pose information of the predetermined landmark point of each 2D image by correcting the 3D head pose information based on the 3D relative coordinates.Type: GrantFiled: May 4, 2020Date of Patent: March 29, 2022Assignee: LG ELECTRONICS INC.Inventors: Taeju Hwang, Beopjung Kim, Namjoon Kim, Yubin Yoon
-
Patent number: 11288828Abstract: An object recognition system based on machine learning includes a 2D image capture module, a 3D space capture module, a data selection module, a data alignment module, a feature capture module and a detection module. The 2D image capture module captures a 2D image. The 3D space capture module captures 3D point cloud data. The data selection module performs object segmentation and depth recognition to select at least one region of interest (ROI). The data alignment module maps coordinates of the at least one ROI to coordinates of the 2D image. The feature capture module calculates features of the 2D image and extracts at least one feature of interest corresponding to the coordinates of the ROI from the features of the 2D image. The detection module obtains the feature of interest and identifies at least one object in the 2D image according to the feature of interest.Type: GrantFiled: January 17, 2020Date of Patent: March 29, 2022Assignee: INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTEInventors: Ming Wu, Kun-Lung Tseng
-
Patent number: 11281934Abstract: In a method for identification of an Internet meme, a plurality of sources is monitored for digital visual content comprising a visual moment and a caption. It is determined whether instances of digital visual content include a same visual moment. Provided the instances of digital visual content include the same visual moment, the instances of digital visual content including the same visual moment are identified as similar digital visual content. Each instance of the similar digital visual content is tracked. Provided a total number of instances of the similar digital visual content exceeds an Internet meme threshold, the similar digital visual content is identified as an Internet meme, wherein the same visual moment is a root visual moment and each caption corresponds to a different iteration of the Internet meme.Type: GrantFiled: February 24, 2020Date of Patent: March 22, 2022Assignee: Gfycat, Inc.Inventors: Jeffrey Harris, Daniel McEleney, Harrison John Dodini, Ernestine Fu
-
Patent number: 11281714Abstract: An image retrieval method is provided. According to an example of the method, based on a first image feature of an image to be retrieved, a binary feature of the image to be retrieved may be determined, and an information entropy code of the image to be retrieved may be further determined based on the binary feature of the image to be retrieved. Then, based on the information entropy code of the image to be retrieved, images similar to the image to be retrieved may be retrieved in a preset image library.Type: GrantFiled: October 18, 2018Date of Patent: March 22, 2022Assignee: Beijing Sankuai Online Technology Co., LtdInventor: Liping Kang
-
Patent number: 11275961Abstract: Provided are character image processing methods and apparatuses, devices, storage medium, and computer programs. The character image processing method mainly comprises: obtaining at least one image block containing a character in a character image to be processed; obtaining image block form transformation information of the image block on the basis of a neural network, the image block form transformation information being used for changing a character orientation in the image block to a predetermined orientation, and the neural network being obtained by means of training using an image block sample having form transformation label information; performing form transformation processing on the character image to be processed according to the image block form transformation information; and performing character recognition on the character image to be processed which is subjected to the form transformation.Type: GrantFiled: November 25, 2019Date of Patent: March 15, 2022Assignee: BEIJING SENSETIME TECHNOLOGY DEVELOPMENT CO., LTD.Inventors: Quan Wang, Ding Liang, Chen Qian
-
Patent number: 11270447Abstract: In a convolutional neural network (CNN) using an encoder-decoder structure for image segmentation, a multi-scale context aggregation module receives an encoded final-stage feature map from the encoder, and sequentially aggregates multi-scale contexts of this feature map from a global scale to a local scale to strengthen semantic relationships of contexts of different scales to improve segmentation accuracy. The multi-scale contexts are obtained by computing atrous convolution on the feature map for different dilation rates. To reduce computation, a channel-wise feature selection (CFS) module is used in the decoder to merge two input feature maps. Each feature map is processed by a global pooling layer followed by a fully connected layer or a 1×1 convolutional layer to select channels of high activation. By subsequent channel-wise multiplication and elementwise summation, only channels with high activation in both feature maps are preserved and enhanced in the merged feature map.Type: GrantFiled: February 10, 2020Date of Patent: March 8, 2022Assignee: Hong Kong Applied Science and Technology Institute Company LimitedInventors: Shangping Liu, Lu Wang, Pingping Zhang, Huchuan Lu
-
Patent number: 11263431Abstract: An automobile identification system utilizes fingerprint sensors disposed an exterior and interior of an automobile to obtain fingerprint images. By means of the reference fingerprint image stored in the database and the determination of the processor, the automobile identification system determines whether the person who attempts to enter the automobile is a driver or not. The automobile identification system does not need the conventional key to open the door of the automobile and activate the engine of the automobile.Type: GrantFiled: August 6, 2019Date of Patent: March 1, 2022Assignee: Ability Opto-Electronics Technology Co., Ltd.Inventors: Yeong-Ming Chang, Chien-Hsun Lai, Shu-Han Chen
-
Patent number: 11250299Abstract: A method is provided for determining entailment between an input premise and an input hypothesis of different modalities. The method includes extracting features from the input hypothesis and an entirety of and regions of interest in the input premise. The method further includes deriving intra-modal relevant information while suppressing intra-modal irrelevant information, based on intra-modal interactions between elementary ones of the features of the input hypothesis and between elementary ones of the features of the input premise. The method also includes attaching cross-modal relevant information to the features from the input premise to the features from the input hypothesis to form a cross-modal representation, based on cross-modal interactions between pairs of different elementary features from different modalities.Type: GrantFiled: October 30, 2019Date of Patent: February 15, 2022Inventors: Farley Lai, Asim Kadav, Ning Xie
-
Patent number: 11250545Abstract: For denoising in SPECT, such as qSPECT, machine learning is used to relate settings to noise structure. Given the SPECT imaging arrangement for a patient, the machine-learned model estimates the structure of the noise. This noise structure may be used to denoise the reconstructed representation.Type: GrantFiled: September 9, 2019Date of Patent: February 15, 2022Assignee: Siemens Medical Solutions USA, Inc.Inventors: Alexander Hans Vija, Michal Cachovan
-
Patent number: 11244198Abstract: In an approach to processing large high dimensional images in parallel without losing accuracy, one or more computer processors determine a required amount of graphics processing unit memory for an image. The one or more computer processors determine one or more coordinate partitions based on the determined required amount of graphics and one or more characteristics of the image. The one or more computer processors determine a padding size for each of the determined one or more coordinate partitions. The one or more computer processors partition the image based on the one or more determined coordinate partitions and the determined padding size. The one or more computer processors generate a prediction for the partitioned image utilizing a trained model.Type: GrantFiled: November 21, 2019Date of Patent: February 8, 2022Assignee: International Business Machines CorporationInventor: Li Zhang
-
Patent number: 11244194Abstract: An object recognition method performed by an object recognition system includes: setting a first binarization area in a portion of a binarization target area of an image, and a first coefficient calculation area including the first binarization area and having an area wider than the first binarization area by a predetermined ratio; determining a first binarization coefficient on the basis of pixel values included in the first coefficient calculation area; performing binarization on the first binarization area through the determined first binarization coefficient; setting a second binarization area in a portion other than the first binarization area, and a second coefficient calculation area including the second binarization area and having an area wider than the second binarization area; determining a second binarization coefficient on the basis of pixel values included in the second coefficient calculation area; and performing binarization on the second binarization area through the determined second binarizType: GrantFiled: February 25, 2020Date of Patent: February 8, 2022Assignee: Fingram Co., Ltd.Inventors: Young Cheul Wee, Young Hoon Ahn, Yang Seong Jin
-
Patent number: 11232318Abstract: The method for vehicle appearance feature recognition includes: multiple region segmentation results of a target vehicle are obtained from an image to be recognized; global feature data and multiple pieces of region feature data are extracted from the image to be recognized based on the multiple region segmentation results; and the global feature data and the multiple pieces of region feature data are fused to obtain appearance feature data of the target vehicle.Type: GrantFiled: November 8, 2019Date of Patent: January 25, 2022Assignee: BEIJING SENSETIME TECHNOLOGY DEVELOPMENT CO., LTD.Inventors: Shuai Yi, Chongdao Wang, Luming Tang, Junjie Yan, Xiaogang Wang
-
Patent number: 11227683Abstract: Various methods and systems are provided for determining and characterizing features of an anatomical structure from a medical image. In one example, a method comprises acquiring a plurality of medical images over time during an exam, registering the segmented anatomical structure between the plurality of medical images, segmenting an anatomical structure in a one of the plurality of medical images after registering the plurality of medical images, creating and characterizing a reference region of interest (ROI) in each of the plurality of medical images, determining characteristics of the anatomical structure by tracking pixel values of the segmented and registered anatomical structure over time, and outputting the determined characteristics on a display device.Type: GrantFiled: January 23, 2020Date of Patent: January 18, 2022Assignee: GE Precision Healthcare LLCInventors: Vincent Jerome Morard, Ronan Quelever, Laura Dumont, Nicolas Gogin, Nicolas Coustaud, Luc Nicodeme
-
Patent number: 11227166Abstract: A method for evaluating images and in particular for evaluating correspondences of images. The method includes (i) providing correspondences between given first and second images, (ii) providing a quality measure or a plurality of quality measures as attributes for characterizing a particular correspondence, (iii) evaluating and conditionally selecting the correspondences, (iv) providing selected correspondences as an evaluation result, the evaluation of correspondences being based on a combination of attributes and the selection of correspondences being based on a result of the evaluation.Type: GrantFiled: September 3, 2018Date of Patent: January 18, 2022Assignee: Robert Bosch GmbHInventors: Stephan Simon, Arne Zender
-
Patent number: 11222224Abstract: An automatic image synthesizing apparatus including an image search unit configured to search for and extract a frame image having an object to be learned or replaced from a plurality of frame images inputted through a vehicle camera; an instance object detection unit configured to detect an instance object to be learned or replaced from the extracted frame image; an image extraction and object class information detection unit configured to extract mask-segmented objects from the extracted frame image and form a data-structure; a standard image mapping unit configured to map a standard image matching a class value of the mask-segmented object; and a generative adversarial network (GAN) network unit configured to receive the selected standard image and the mask segmentation image and learn a replacement image or produce a replacement image based on a learning result.Type: GrantFiled: February 27, 2020Date of Patent: January 11, 2022Assignee: Hyundai Mobis Co., Ltd.Inventors: Dong Yul Lee, Jin Aeon Lee
-
Patent number: 11216971Abstract: A three-dimensional bounding box is determined from a two-dimensional image and a point cloud. A feature vector associated with the image and a feature vector associated with the point cloud may be passed through a neural network to determine parameters of the three-dimensional bounding box. Feature vectors associated with each of the points in the point cloud may also be determined and considered to produce estimates of the three-dimensional bounding box on a per-point basis.Type: GrantFiled: August 30, 2019Date of Patent: January 4, 2022Assignee: Zoox, Inc.Inventors: Danfei Xu, Dragomir Dimitrov Anguelov, Ashesh Jain
-
Patent number: 11216981Abstract: In one aspect, a system for calibrating image data during the performance of an agricultural operation may include an imaging device configured to capture image data associated with a portion of a field within a field of view of the imaging device. Additionally, the system may include a color indicator positioned within the field of view of the imaging device. Moreover, the system may include a controller configured to receive the image data from the imaging device and determine an apparent color value of the color indicator based on the received image data. Furthermore, the controller may be configured to adjust at least one of an operating parameter of the imaging device or a variable associated with processing the received image data based on the determined apparent color value.Type: GrantFiled: July 26, 2019Date of Patent: January 4, 2022Assignee: CNH Industrial America LLCInventors: Luca Ferrari, Trevor Stanhope