专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

11. 发明申请

WO2021146234A1 SHIFT INVARIANT LOSS FOR DEEP LEARNING BASED IMAGE SEGMENTATION 审中-公开
公开(公告)号：WO2021146234A1
公开(公告)日：2021-07-22
申请号：PCT/US2021/013172
申请日：2021-01-13
申请人： VERILY LIFE SCIENCES LLC
发明人： WU, Cheng-Hsun , BEHROOZ, Ali
IPC分类号： G06K9/00 , G06K9/62 , G06F17/15 , G06F17/18 , G06K9/0014 , G06K9/6203 , G06K9/6206 , G06N3/08 , G06T2207/20132 , G06T7/11 , G10L15/04 , G10L15/16
摘要： Systems and methods of improving alignment in dense prediction neural networks are disclosed. A method includes identifying, at a computing system, an input data set and a label data set with one or more first parts of the input data set corresponding to a label. The computing system processes the input data set using a neural network to generate a predicted label data set that identifies one or more second parts of the input data set predicted to correspond to the label. The computing system determines an alignment result using the predicted label data set and the label data set and a transformation of the one or more first parts, including a shift, rotation, scaling, and/or deformation, based on the alignment result. The computing system computes a loss score using the transformation, label data and the predicted label data set and updates the neural network based on the loss score.

12. 发明申请

WO2021082188A1 全球导航卫星系统实时钟差评估算法审中-公开
公开(公告)号：WO2021082188A1
公开(公告)日：2021-05-06
申请号：PCT/CN2019/123836
申请日：2019-12-07
申请人：中海北斗深圳导航技术有限公司
发明人：陈国 , 赵齐乐 , 陈卓敏
IPC分类号： G06F17/11 , G06F17/15 , G06F17/16 , G01S19/20
摘要：一种GNSS实时钟差的评估算法，假设能够获取GNSS实时轨道和钟差以及参考轨道和钟差产品，且解算的GNSS实时钟差和参考产品含有共同的卫星钟差数据，利用实时轨道和参考轨道产品计算钟差一致性改正项，利用中位数粗差探测法实现对GNSS实时钟差的异常数据探测和标记，利用全变差正则化方法实现对GNSS钟差的单差序列进行跳变探测和分段标记，计算跳变大小；采用中位数粗差探测方法，可以快速有效地识别和剔除实时钟差中的异常数据，防止异常数据对实时钟差评估结果的损害；针对由于具体实时钟差数据策略(如数据中断、基准钟的切换等)造成的钟差频繁跳变问题，创造性的利用全变差正则化方法实现对实时钟差的跳变探测。

13. 发明申请

WO2021080724A1 THREE DIMENSIONAL CONVOLUTION IN NEURAL NETWORK PROCESSOR 审中-公开
公开(公告)号：WO2021080724A1
公开(公告)日：2021-04-29
申请号：PCT/US2020/052053
申请日：2020-09-22
申请人： APPLE INC.
发明人： MILLS, Christopher, L. , PARK, Sung, Hee
IPC分类号： G06F17/15 , G06F9/30 , G06N3/04 , G06T1/20
摘要： Embodiments of the present disclosure relate to a neural engine of a neural processor circuit having multiple multiply-add circuits and an accumulator circuit coupled to the multiply-add circuits. The multiply-add circuits perform multiply-add operations of a three dimensional convolution on a work unit of input data using a kernel to generate at least a portion of output data in a processing cycle. The accumulator circuit includes multiple batches of accumulators. Each batch of accumulators receives and stores, after the processing cycle, the portion of the output data for each output depth plane of multiple output depth planes. A corresponding batch of accumulators stores, after the processing cycle, the portion of the output data for a subset of the output channels and for each output depth plane.

14. 发明申请

WO2021063170A1 一种基于球冠函数的双轴旋转系统指向误差的修正方法审中-公开
公开(公告)号：WO2021063170A1
公开(公告)日：2021-04-08
申请号：PCT/CN2020/115069
申请日：2020-09-14
申请人：中国科学院国家天文台南京天文光学技术研究所
发明人：郑奕 , 李颖 , 吴兆祥 , 梁斌 , 汤丽峰
IPC分类号： G05D3/12 , G06F17/15
摘要：一种基于球冠函数的双轴旋转系统指向误差的修正方法，包括：误差采集：在星表中选取全天均匀分布的恒星或射电源进行跟踪观测，得到恒星的理论位置和测量位置，将测量位置和理论位置相减，得到误差分布；误差模型拟合：将得到的误差分布，选择合适的正交球冠函数模型，进行拟合，计算出误差拟合系数；其中，正交球冠函数模型包括半球谐函数HSH、Zernike球冠函数ZSF和纵向球冠函数LSF；误差控制补偿：把误差模型和相关的拟合系数，放入指向控制系统，进行补偿。采用基于球冠函数的双轴旋转系统指向误差的修正方法，模型稳定性强，不易受到测量噪声的影响；不需要需根据望远镜的机架形式，确定模型的形式，修正精度高。

15. 发明申请

WO2021048627A2 SYSTEMS AND METHODS FOR CONTEXT-BASED CONTENT GENERATION 审中-公开
公开(公告)号：WO2021048627A2
公开(公告)日：2021-03-18
申请号：PCT/IB2020/000752
申请日：2020-09-11
申请人： ADEPTION LIMITED
发明人： SANDER-EDWARDS, Carl , DESHPANDE, Amol
IPC分类号： H04N21/236 , G06F16/48 , G06F17/15 , G06F17/16 , H04N21/442 , H04N21/45 , H04N21/47 , G06F16/437 , G06F16/45 , G06F16/483 , G06N5/04
摘要： This invention discloses a computer-implemented method, caused by a server, for hierarchical causality-based stitching of content and for serving said stitched content as output content, said method comprising: tracking, and measuring, a first set of markers, for a first content consumer, consuming a first content item; tracking, and measuring, a first set of markers, for a second content consumer, consuming a second content item; receiving, by said first content consumer, a request corresponding to a marker from said first set of markers; computing a "pertinence indicator"; computing a "colliding score"; automatically collating said first content item, correlative to said first user, and a second content item, correlative to said second user, to form at least an output content, if said "pertinence indicator" is within said pre-defined rules of correlation and if said "colliding score" is within said pre-determined threshold; and serving said collated output content.

16. 发明申请

WO2021035057A8 GENERAL PADDING SUPPORT FOR CONVOLUTION ON SYSTOLIC ARRAYS 审中-公开
公开(公告)号：WO2021035057A8
公开(公告)日：2021-02-25
申请号：PCT/US2020/047215
申请日：2020-08-20
申请人： GOOGLE LLC
发明人： MAJNEMER, David Alexander , HECHTMAN, Blake Alan , ROUNE, Bjarke Hammersholt
IPC分类号： G06N3/04 , G06N3/063 , G06F17/15 , G06F17/16 , G06N3/0454
摘要： Methods and systems, including computer programs encoded on a computer storage medium. In one aspect, a method includes the actions of receiving a request to perform convolutional computations for a neural network on a hardware circuit having a matrix computation unit, the request specifying the convolutional computation to be performed on a feature tensor and a filter and padding applied to the feature tensor prior to performing the convolutional computation; and generating instructions that when executed by the hardware circuit cause the hardware circuit to perform operations comprising: transferring feature tensor data from a main memory of the hardware circuit to a scratchpad memory of the hardware circuit; and repeatedly performing the following operations: identifying a current subset of the feature tensor; and determining whether a memory view into the scratchpad memory for the current subset is consistent with a memory view of the current subset in the main memory.

17. 发明申请

WO2020176841A1 SYSTEMS AND METHODS FOR PREDICTING THE GEOMETRY AND INTERNAL STRUCTURE OF TURBINE BLADES 审中-公开
公开(公告)号：WO2020176841A1
公开(公告)日：2020-09-03
申请号：PCT/US2020/020330
申请日：2020-02-28
申请人： GEORGIA TECH RESEARCH CORPORATION
发明人： GENTRY, Russell , TASISTRO-HART, Benjamin
IPC分类号： F01D5/14 , G06T17/20 , G06F17/15
摘要： An exemplary embodiment provides a method of predicting a geometry of a turbine blade, comprising: obtaining a point cloud of the turbine blade, the point cloud comprising a plurality of points corresponding to locations on an external contour of the turbine blade; generating an external mesh for the turbine blade based on point cloud; defining a reference axis along the length of the blade; generating a plurality of sampling planes along the length of the reference axis, each sampling plane being normal to the reference axis and having a shape defined by portions of the mesh intersecting the sampling plane; for each of the plurality of sampling planes, matching the sampling plane to an airfoil profile in a database of known airfoil profiles; and creating a geometrical representation of the turbine blade by placing the airfoils on the reference axis and connecting the perimeter of each of the airfoils.

18. 发明申请

WO2020116745A1 고속 딥러닝을 위한 컨볼루션 방법 审中-公开
公开(公告)号：WO2020116745A1
公开(公告)日：2020-06-11
申请号：PCT/KR2019/010624
申请日：2019-08-21
申请人： 아주대학교 산학협력단
发明人： 선우명훈 , 김태선
IPC分类号： G06N3/08 , G06F17/15
摘要： 본 발명에 따른 고속 딥러닝을 위한 컨볼루션 방법은 (a) 상기 컨볼루션 시스템의 피쳐 맵 수신부가 N개의 채널로 구성된 피쳐 맵을 수신하는 단계; (b) 상기 컨볼루션 시스템의 주제어부가 N개의 채널로 구성된 피쳐 맵에서 `0`번째 채널을 선택하는 단계; (c) 상기 주제어부가 `0`번째 채널의 피쳐 맵에서 x, y 좌표가 `0`인 좌표를 확인하는 단계; (d) 상기 컨볼루션 시스템의 컨볼루션 계산부가 상기 피쳐 맵에 대해 가로방향 및 세로방향으로 2만큼 이동해가며 컨볼루션 연산과 ReLU(Rectified Liner Unit)연산을 수행하는 코스(coarse)단계; (e) 상기 컨볼루션 시스템의 채널 변경부가 `0`번째 채널의 피쳐 맵에 대해 코스(coarse)단계가 완료되면 다음 채널로 채널을 변경하는 단계; (f) 상기 주제어부가 변경된 채널이 N보다 큰지 작은지 판단하는 단계; 및 (g) 상기 주제어부가 상기 (f)단계에서 변경된 채널이 N보다 큰 경우 모든 채널에 대한 컨볼루션 연산이 완료된 것으로 판단하고 피쳐 맵 출력부를 통해 피쳐맵을 출력하는 단계;를 포함하여 컨볼루션 신경망에서 대부분을 차지하는 컨볼루션 연산을 줄여 딥러닝에서의 추론 속도를 높일 수 있는 효과가 있다.

19. 发明申请

WO2020087742A1 用于实现卷积运算的处理元件、装置和方法审中-公开
公开(公告)号：WO2020087742A1
公开(公告)日：2020-05-07
申请号：PCT/CN2018/124828
申请日：2018-12-28
申请人：深圳云天励飞技术有限公司
发明人：黎立煌 , 李炜 , 曹庆新
IPC分类号： G06F17/15 , G06N3/04
摘要：一种用于实现卷积运算的处理元件、装置和方法。其中，处理元件(PE)包括：第一缓冲器(11)，被配置为存储输入数据以及与卷积运算对应的权值；移位单元(12)，被配置为对输入数据执行移位操作，生成第一中间数据；多个操作单元(13)，被配置为基于权值和第一中间数据执行二维卷积运算中的至少一部分运算，并生成输出数据。能进行二维卷积运算且能提高数据的重用率，还能最大限度地利用和挖掘硬件的并行性，并能降低硬件的功耗。

20. 发明申请

WO2020069505A1 REAL-TIME GUIDANCE OF RADIOFREQUENCY ABLATION CATHETER CONTACT ORIENTATION WITH CARDIAC TISSUE USING OPTICAL COHERENCE TOMOGRAPHY 审中-公开
公开(公告)号：WO2020069505A1
公开(公告)日：2020-04-02
申请号：PCT/US2019/053817
申请日：2019-09-30
申请人： THE TRUSTEES OF COLUMBIA UNIVERSITY IN THE CITY OF NEW YORK
发明人： HENDON, Christine , SINGH-MOON, Rajinder , YU, Xin
IPC分类号： G06T7/00 , G06N3/08 , A61B3/10 , G06F17/15 , G06N3/04 , A61B6/02
摘要： [0056] An exemplary system, method and computer-accessible medium for determining particular for a portion(s) of an organ of a patient(s) can be provided, which can include, for example, receiving an in vivo optical coherence tomography (OCT) imaging information for the portion(s), and determining (i) a lesion formation on or in the portion(s), (ii) a contact of a catheter on or with the portion(s), or (iii) an angle of the catheter with respect to the portion(s), by applying a convolutional neural network(s) to the in vivo OCT imaging information.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式