-
公开(公告)号：US20220122639A1
公开(公告)日：2022-04-21
申请号：US17492781
申请日：2021-10-04
Applicant: Amazon Technologies, Inc.
Inventor： Matthew Alan Townsend , Rohith Mysore Vijaya Kumar , Yadunandana Nagaraja Rao , Ambrish Tyagi , Eduard Oks , Apoorv Chaudhri
IPC: G11B27/031 , H04N9/79
Abstract: Devices, systems and methods are disclosed for improving story assembly and video summarization. For example, video clips may be received and a theme may be determined from the received video clips based on annotation data or other characteristics of the received video data. Individual moments may be extracted from the video clips, based on the selected theme and the annotation data. The moments may be ranked based on a priority metric corresponding to content determined to be desirable for purposes of video summarization. Select moments may be chosen based on the priority metric and a structure may be determined based on the selected theme. Finally, a video summarization may be generated using the selected theme and the structure, the video summarization including the select moments.
-
公开(公告)号：US10909349B1
公开(公告)日：2021-02-02
申请号：US16450499
申请日：2019-06-24
Applicant: Amazon Technologies, Inc.
Inventor： Shashank Tripathi , Visesh Chari , Ambrish Tyagi , Amit Kumar Agrawal , James Rehg , Siddhartha Chandra
Abstract: Techniques are generally described for object detection in image data. First image data comprising a three-dimensional model representing an object may be received. First background image data comprising a first plurality of pixel values may be received. A first feature vector representing the three-dimensional model may be generated. A second feature vector representing the first plurality of pixel values of the first background image data may be generated. A first machine learning model may generate a transformed representation of the three-dimensional model using the first feature vector. First foreground image data comprising a two-dimensional representation of the transformed representation of the three-dimensional model may be generated. A frame of composite image data may be generated by combining the first foreground image data with the first background image data.
-
公开(公告)号：US10482925B1
公开(公告)日：2019-11-19
申请号：US15783584
申请日：2017-10-13
Applicant: Amazon Technologies, Inc.
Inventor： Ambrish Tyagi , Suresh Bholabhai Lakhani , Rohith Mysore Vijaya Kumar , Yadunandana Nagaraja Rao , Amit Kumar Agrawal
Abstract: A system and method for selecting portions of video data from preview video data is provided. The system may extract image features from the preview video data and discard video frames associated with poor image quality based on the image features. The system may determine similarity scores between individual video frames and corresponding transition costs and may identify transition points in the preview video data based on the similarity scores and/or transition costs. The system may select portions of the video data for further processing based on the transition points and the image features. By selecting portions of the video data, the system may reduce a bandwidth consumption, processing burden and/or latency associated with uploading the video data or performing further processing.
-
公开(公告)号：US09436870B1
公开(公告)日：2016-09-06
申请号：US14298745
申请日：2014-06-06
Applicant: Amazon Technologies, Inc.
Inventor： Stephen Vincent Mangiat , Ambrish Tyagi
CPC classification number: G06K9/00221 , G06F3/005 , G06F3/012 , G06F3/017 , G06K9/00228 , G06K9/00362 , G06K9/2027
Abstract: The subject technology provides embodiments for tracking a user's face/head (or another object) using one or more cameras provided by a computing device. Embodiments implement exposure sweeping based on an average intensity of a current scene to a target intensity for a given image. If a face is not detected, an exposure duration and/or gain may be adjusted and the face detection is performed again. Once the face is detected, an average intensity of a virtual bounding box surrounding the detected face is determined and exposure sweeping may be performed solely within the virtual bounding box to reach a target intensity. When the average intensity is within a predetermined threshold of the target intensity, the detected face may be at an optimal exposure. Embodiments also provide for switching to another camera(s) of the computing device when not detecting a face in the image upon performing a full exposure sweep.
Abstract translation: 主题技术提供使用由计算设备提供的一个或多个照相机跟踪用户的脸部/头部（或另一个对象）的实施例。 实施例将基于当前场景的平均强度的曝光扫描实现为给定图像的目标强度。 如果未检测到脸部，则可以调整曝光持续时间和/或增益，并再次执行脸部检测。 一旦检测到脸部，确定围绕检测到的脸部的虚拟边界框的平均强度，并且可以在虚拟边界框内单独执行曝光扫描以达到目标强度。 当平均强度在目标强度的预定阈值内时，检测到的面可以处于最佳曝光。 实施例还提供在执行完全曝光扫描时在不检测图像中的面部时切换到计算设备的另一个相机。
-
公开(公告)号：US20150362989A1
公开(公告)日：2015-12-17
申请号：US14307483
申请日：2014-06-17
Applicant: Amazon Technologies, Inc.
Inventor： Ambrish Tyagi , Kah Kuen Fu , Tianyang Ma , Kenneth Mark Karakotsios , Michael Lee Sandige , David Wayne Stafford , Steven Bennett
CPC classification number: G06F3/011 , G06F3/017 , G06K9/00255 , G06K9/00268 , G06K9/2018 , G06K9/2027 , G06K9/209 , G06K9/6201 , G06K9/6256 , G06K2009/00328 , H04N5/33 , H04N7/18
Abstract: Object tracking, such as may involve face tracking, can utilize different detection templates that can be trained using different data. A computing device can determine state information, such as the orientation of the device, an active illumination, or an active camera to select an appropriate template for detecting an object, such as a face, in a captured image. Information about the object, such as the age range or gender of a person, can also be used, if available, to select an appropriate template. In some embodiments instances of templates can be used to process various orientations, while in other embodiments specific orientations, such as upside down orientations, may not be processed for reasons such as rate of inaccuracies or infrequency of use for the corresponding additional resource overhead.
Abstract translation: 对象跟踪，例如可能涉及面部跟踪，可以利用可以使用不同数据进行训练的不同检测模板。 计算设备可以确定状态信息，诸如设备的方向，活动照明或活动照相机，以选择用于检测拍摄图像中的对象（例如脸部）的适当模板。 关于对象的信息，如人的年龄范围或性别，也可以使用（如果有的话）来选择适当的模板。 在一些实施例中，可以使用模板的实例来处理各种取向，而在其他实施例中，由于诸如对于相应的额外资源开销的使用的不准确率或不频率的原因，可能无法处理特定取向，例如上下取向。
-
Patent Agency Ranking
公开(公告)号：US11526697B1
公开(公告)日：2022-12-13
申请号：US16814526
申请日：2020-03-10
Applicant: Amazon Technologies, Inc.
Inventor： Shashank Tripathi , Ambrish Tyagi , Amit Kumar Agrawal , Siddhant Ranade
Abstract: Devices and techniques are generally described for estimating three-dimensional pose data. In some examples, a first machine learning network may generate first three-dimensional (3D) data representing input 2D data. In various examples, a first 2D projection of the first 3D data may be generated. A determination may be made that the first 2D projection conforms to a distribution of natural 2D data. A second machine learning network may generate parameters of a 3D model based at least in part on the input 2D data and based at least in part on the first 3D data. In some examples, second 3D data may be generated using the parameters of the 3D model.
-
公开(公告)号：US10860836B1
公开(公告)日：2020-12-08
申请号：US16192433
申请日：2018-11-15
Applicant: Amazon Technologies, Inc.
Inventor： Ambrish Tyagi , Amit Kumar Agrawal , Siddhartha Chandra , Visesh Uday Kumar Chari , Shashank Tripathi , James Rehg
Abstract: Techniques are generally described for object detection in image data. First image data comprising a first plurality of pixel values representing an object and a second plurality of pixel values representing a background may be received. First foreground image data and first background image data may be generated from the first image data. A first feature vector representing the first plurality of pixel values may be generated. A second feature vector representing a first plurality of pixel values of second background image data may be generated. A first machine learning model may determine a first operation to perform on the first foreground image data. A transformed representation of the first foreground image data may be generated by performing the first operation on the first foreground image data. Composite image data may be generated by compositing the transformed representation of the first foreground image data with the second background image data.
-
公开(公告)号：US10007860B1
公开(公告)日：2018-06-26
申请号：US14976696
申请日：2015-12-21
Applicant: Amazon Technologies, Inc.
Inventor： David Allen Fotland , Ambrish Tyagi
CPC classification number: G06K9/3233 , G06K9/00369 , G06T7/11 , G06T7/187 , G06T7/194 , G06T2207/20132
Abstract: The techniques described herein may identify images that likely depict one or more items by comparing features of the items to features of different regions-of-interest (ROIs) of the images. For instance, some of the images may include a user, and the techniques may define multiple regions within the image corresponding to different portions of the user. The techniques may then use a trained convolutional neural network or any other type of trained classifier to determine, for each region of the image, whether the region depicts a particular item. If so, the techniques may designate the corresponding image as depicting the item and may output an indication that the image depicts the item. The techniques may perform this process for multiple images, outputting an indication of each image deemed to depict the particular item.
-
公开(公告)号：US09953242B1
公开(公告)日：2018-04-24
申请号：US14976850
申请日：2015-12-21
Applicant: Amazon Technologies, Inc.
Inventor： Ambrish Tyagi , David Allen Fotland
CPC classification number: G06K9/6212 , G06K9/38 , G06K9/4652 , G06K9/52 , G06K2009/4666 , G06K2009/6213 , G06T7/0059 , G06T7/0081 , G06T2207/10024 , G06T2207/20132 , G06T2207/30196
Abstract: The techniques described herein may identify images that likely depict one or more items by comparing features of the items to features of different regions-of-interest (ROIs) of the images. When a user requests to identify images that depict a particular item, the techniques may determine a region-of-interest (ROI) size based on the size of the requested item. The techniques may then search multiple images using the ROI size.
-
公开(公告)号：US09824455B1
公开(公告)日：2017-11-21
申请号：US14976555
申请日：2015-12-21
Applicant: Amazon Technologies, Inc.
Inventor： Abhishek Singh , Rohith Mysore Vijaya Kumar , Ambrish Tyagi
CPC classification number: G06T7/0081 , G06T7/0097 , G06T7/204 , G06T7/215 , G06T2207/10016 , G06T2207/20076 , G06T2207/20144 , G06T2207/20164 , G06T2207/30241
Abstract: Various examples are directed to systems and methods for determining foreground regions in video frames. A computing device may select from a first frame of a video, a plurality of scene point locations and divide the first frame into a plurality of sections. For a first section of the first frame, the computing device may generate a first vector subspace, basis vectors of the first vector subspace are trajectories of scene point locations in the first section. The computing device may determine that a projection error for a first scene point location in the first section is greater than a projection error threshold and write an indication of the first pixel value to a listing of foreground pixel values.
-
-
-
-
-
-
-
-
-
Search Results
32
records
(took 0.045 seconds)
