知搜-全球专利检索

  1. 登录
  2. 注册

搜索结果

24 条记录 (耗时 0.018 秒)

    Reducing power consumption in a neural network environment using data management
    4.
    发明授权
    Reducing power consumption in a neural network environment using data management 有权

    公开(公告)号：US10996739B2

    公开(公告)日：2021-05-04

    申请号：US15847785

    申请日：2017-12-19

    申请人： Microsoft Technology Licensing, LLC

    发明人： Amol Ashok Ambardekar ,  Chad Balling McBride ,  George Petre ,  Kent D. Cedola ,  Larry Marvin Wall

    IPC分类号： G06N3/04 ,  G06N3/08 ,  G06N3/063 ,  G06F17/16 ,  G06F1/32 ,  G06F1/3234

    摘要： Techniques to provide for improved (i.e., reduced) power consumption in an exemplary neural network (NN) and/or Deep Neural Network (DNN) environment using data management. Improved power consumption in the NN/DNN may be achieved by reducing a number of bit flips needed to process operands associated with one or more storages. Reducing the number bit flips associated with the NN/DNN may be achieved by multiplying an operand associated with a first storage with a plurality of individual operands associated with a plurality of kernels of the NN/DNN. The operand associated with the first storage may be neuron input data and the plurality of individual operands associated with the second storage may be weight values for multiplication with the neuron input data. The plurality of kernels may be arranged or sorted and subsequently processed in a manner that improves power consumption in the NN/DNN.

    Compression-encoding scheduled inputs for matrix computations
    6.
    发明授权
    Compression-encoding scheduled inputs for matrix computations 有权

    公开(公告)号：US11526581B2

    公开(公告)日：2022-12-13

    申请号：US17085337

    申请日：2020-10-30

    申请人： Microsoft Technology Licensing, LLC

    发明人： Shuayb M. Zarar ,  Amol Ashok Ambardekar ,  Jun Zhang

    IPC分类号： G06F17/16

    摘要： A method of performing matrix computations includes receiving a compression-encoded matrix including a plurality of rows. Each row of the compression-encoded matrix has a plurality of defined element values and, for each such defined element value, a schedule tag indicating a schedule for using the defined element value in a scheduled matrix computation. The method further includes loading the plurality of rows of the compression-encoded matrix into a corresponding plurality of work memory banks, and providing decoded input data to a matrix computation module configured for performing the scheduled matrix computation. For each work memory bank, a next defined element value and a corresponding schedule tag are read. If the schedule tag meets a scheduling condition, the next defined element value is provided to the matrix computation module. Otherwise, a default element value is provided to the matrix computation module.

    Aligning input image data with model input data to generate image annotations
    10.
    发明授权
    Aligning input image data with model input data to generate image annotations 有权

    公开(公告)号：US11514648B2

    公开(公告)日：2022-11-29

    申请号：US17133493

    申请日：2020-12-23

    申请人： MICROSOFT TECHNOLOGY LICENSING, LLC

    发明人： Harpreet Singh Sawhney ,  Ning Xu ,  Amol Ashok Ambardekar ,  Moses Obadeji Olafenwa

    IPC分类号： G06T19/00 ,  G06T7/70 ,  G06N3/04 ,  G06N3/08 ,  G06T17/00

    摘要： An image data annotation system automatically annotates a physical object within individual images frames of an image sequence with relevant object annotations based on a three-dimensional (3D) model of the physical object. Annotating the individual image frames with object annotations includes updating individual image frames within image input data to generate annotated image data that is suitable for reliably training a DNN object detection architecture. Exemplary object annotations that the image data annotation system can automatically apply to individual image frames include, inter alia, object pose, image pose, object masks, 3D bounding boxes composited over the physical object, 2D bounding boxes composited over the physical object, and/or depth map information. Annotating the individual image frames may be accomplished by aligning the 3D model of the physical object with a multi-view reconstruction of the physical object that is generated by inputting an image sequence into a Structure-from-Motion and/or Multi-view Stereo pipeline.