知搜-全球专利检索

  1. 登录
  2. 注册

搜索结果

38 条记录 (耗时 0.023 秒)

    Determining layer ranks for compression of deep networks
    1.
    发明授权
    Determining layer ranks for compression of deep networks 有权

    公开(公告)号：US11586924B2

    公开(公告)日：2023-02-21

    申请号：US15877723

    申请日：2018-01-23

    申请人： QUALCOMM Incorporated

    发明人： Anthony Sarah ,  Raghuraman Krishnamoorthi

    IPC分类号： G06N3/08 ,  G06N3/084 ,  G06N3/04

    摘要： An apparatus of operating a computational network is configured to determine a low-rank approximation for one or more layers of the computational network based at least in part on a set of residual targets. A set of candidate rank vectors corresponding to the set of residual targets may be determined. Each of the candidate rank vectors may be evaluated using an objective function. A candidate rank vector may be selected and used to determine the low rank approximation. The computational network may be compressed based on the low-rank approximation. In turn the computational network may be operated using the one or more compressed layers.

    INTERACTIVE AND AUTOMATIC 3-D OBJECT SCANNING METHOD FOR THE PURPOSE OF DATABASE CREATION
    3.
    发明申请
    INTERACTIVE AND AUTOMATIC 3-D OBJECT SCANNING METHOD FOR THE PURPOSE OF DATABASE CREATION 有权
    用于数据库创建目的的交互式和自动三维对象扫描方法

    公开(公告)号：US20140363048A1

    公开(公告)日：2014-12-11

    申请号：US14302056

    申请日：2014-06-11

    申请人： QUALCOMM Incorporated

    发明人： Bojan Vrcelj ,  Daniel Knoblauch ,  Raghuraman Krishnamoorthi ,  Serafin Diaz Spindola

    IPC分类号： G06K9/78 ,  G06K9/62

    CPC分类号： G06K9/78 ,  G06K9/6228 ,  G06T7/75 ,  G06T2207/10028 ,  G06T2207/20081

    摘要： Systems, methods, and devices are described for capturing compact representations of three-dimensional objects suitable for offline object detection, and storing the compact representations as object representation in a database. One embodiment may include capturing frames of a scene, identifying points of interest from different key frames of the scene, using the points of interest to create associated three-dimensional key points, and storing key points associated with the object as an object representation in an object detection database.

    摘要翻译： 描述了用于捕获适合于离线对象检测的三维对象的紧凑表示以及将紧凑表示作为对象表示存储在数据库中的系统，方法和设备。 一个实施例可以包括捕获场景的帧，使用感兴趣点来识别来自场景的不同关键帧的兴趣点，以创建相关联的三维关键点，以及将与对象相关联的关键点存储为对象表示。 对象检测数据库。

    DETERMINING CAPACITY OF SEARCH STRUCTURES
    4.
    发明申请
    DETERMINING CAPACITY OF SEARCH STRUCTURES 有权
    确定搜索结构的能力

    公开(公告)号：US20140059037A1

    公开(公告)日：2014-02-27

    申请号：US13802082

    申请日：2013-03-13

    申请人： QUALCOMM INCORPORATED

    发明人： Ashwin Swaminathan ,  Raghuraman Krishnamoorthi ,  Serafin Diaz Spindola ,  Murali Ramaswamy Chari

    IPC分类号： G06F17/30

    CPC分类号： G06F17/3028 ,  G06F17/30247 ,  G06F17/30271 ,  G06F17/30961

    摘要： A system for storing target images for object recognition predicts a querying performance for the target image if the target image were included in a search tree of a database. The search tree has a universal search tree structure that is fixed so that it does not change with the addition of new target images. The target image is selected for inclusion or exclusion in the search tree based on the based on the querying performance, wherein the fixed tree structure of the search tree does not change if inclusion of the target image is selected.

    摘要翻译： 如果目标图像被包括在数据库的搜索树中，则用于存储用于对象识别的目标图像的系统预测目标图像的查询性能。 搜索树具有固定的通用搜索树结构，使得它不随着添加新的目标图像而改变。 基于查询性能，在搜索树中选择目标图像以包含或排除，其中如果选择了包含目标图像，搜索树的固定树结构不改变。