公开(公告)号：US20190303683A1

公开(公告)日：2019-10-03

申请号：US16232780

申请日：2018-12-26

Applicant: Amazon Technologies, Inc.

Inventor： Mark Eugene Pearson ,  Dynin Hong Khem ,  Peter Van Tuyl Bentley ,  William Christopher Banta ,  Kevin Michael Gordon ,  Manlio Armando Lo Conte

IPC: G06K9/00 ,  G11B27/30 ,  G11B27/02 ,  G06T7/20 ,  G06T7/60

Abstract: Devices, systems and methods are disclosed for improving a playback of video data and generation of a video summary. For example, annotation data may be generated for individual video frames included in the video data to indicate content present in the individual video frames, such as faces, objects, pets, speech or the like. A video summary may be determined by calculating a priority metric for individual video frames based on the annotation data. In response to input indicating a face and a period of time, a video summary can be generated including video segments focused on the face within the period of time. The video summary may be directed to multiple faces and/or objects based on the annotation data.

公开(公告)号：US10169659B1

公开(公告)日：2019-01-01

申请号：US14863705

申请日：2015-09-24

Applicant: Amazon Technologies, Inc.

Inventor： Mark Eugene Pearson ,  Dynin Hong Khem ,  Peter Van Tuyl Bentley ,  William Christopher Banta ,  Kevin Michael Gordon ,  Manlio Armando Lo Conte

IPC: H04N5/93 ,  G06K9/00 ,  G06K9/52 ,  G06T7/60 ,  G06T7/20 ,  G11B27/02 ,  G11B27/30 ,  G10L25/78 ,  G10L25/81

Abstract: Devices, systems and methods are disclosed for improving a playback of video data and generation of a video summary. For example, annotation data may be generated for individual video frames included in the video data to indicate content present in the individual video frames, such as faces, objects, pets, speech or the like. A video summary may be determined by calculating a priority metric for individual video frames based on the annotation data. In response to input indicating a face and a period of time, a video summary can be generated including video segments focused on the face within the period of time. The video summary may be directed to multiple faces and/or objects based on the annotation data.