Adaptive retrieval of objects from remote storage

    公开(公告)号:US11474948B2

    公开(公告)日:2022-10-18

    申请号:US16798103

    申请日:2020-02-21

    申请人: NETFLIX, INC.

    摘要: Various embodiments of the present application set forth a computer-implemented method for accessing data comprising identifying a first set of read operations occurring during a first time period, where each read operation included in the set of read operations is associated with retrieving a different portion of at least one object from a storage system, determining a byte density associated with the set of read operations, where the byte density indicates a size of contiguous portions of the at least one object that were retrieved during the first time period, and determining, based on the byte density, a pre-buffering block size for a read operation during a second period, where the pre-buffering block size specifies a size of a portion of at least one object that is to be retrieved from the storage system.

    Systems and methods for implementing shared locks to stream telemetry data between server instances

    公开(公告)号:US11457086B1

    公开(公告)日:2022-09-27

    申请号:US17313835

    申请日:2021-05-06

    申请人: Netflix, Inc.

    摘要: The disclosed computer-implemented method includes identifying a target networking device communicatively connected to a cluster of networked server instances. The networked server instances are configured to consume network telemetry messages. The method further includes applying a shared lock to the target networking device, so that only a single networked server instance has access to network telemetry messages produced by the target networking device. The method also includes replicating the network telemetry messages produced by the locked target networking device to other server instances in the cluster, and then diverting requests for network telemetry messages from the locked target networking device to another server instance in the cluster of networked server instances. This allows the other server instance in the cluster to service the diverted telemetry data requests for the locked target networking device using the replicated network telemetry messages. Various other methods, systems, and computer-readable media are also disclosed.

    TECHNIQUES FOR AUTOMATICALLY EXTRACTING COMPELLING PORTIONS OF A MEDIA CONTENT ITEM

    公开(公告)号:US20220277564A1

    公开(公告)日:2022-09-01

    申请号:US17745718

    申请日:2022-05-16

    申请人: NETFLIX, INC.

    IPC分类号: G06V20/40 G06K9/62 G06V40/16

    摘要: In various embodiments, a clip application computes a set of appearance values for an appearance metric based on shot sequences associated with a media content item. Each appearance value in the set of appearance values indicates a prevalence of a first character in a different shot sequence associated with the media content item. The clip application then performs one or more clustering operations on the shot sequences based on the set of appearance values to generate a first shot cluster. Subsequently, the clip application generates a clip for the media content item based on the first shot cluster. The clip application transmits the clip to an endpoint device for display. Advantageously, relative to primarily manual approaches, the clip application can more efficiently and reliably generate clips for media content items.

    SYSTEMS AND METHODS FOR PREPROCESSING APPLICATION FUNCTIONS FOR FASTER STARTUP

    公开(公告)号:US20220222082A1

    公开(公告)日:2022-07-14

    申请号:US17145023

    申请日:2021-01-08

    申请人: Netflix, Inc.

    IPC分类号: G06F9/445 G06N20/00

    摘要: The disclosed computer-implemented method may include predicting, by a machine learning model, a user action and a timing of the user action for an application on a computing device. The method may also include determining that an expected lag in executing the user action is greater than a predetermined threshold based on one or more resource constraints of the computing device and initializing a setup of the application to reduce the expected lag in executing the user action based on the predicted timing. Furthermore, the method may include prefetching one or more application components in response to initializing the setup of the application and preprocessing at least a portion of a function of the application used to execute the user action. Finally, the method may include executing the user action for the application in response to a user request. Various other methods, systems, and computer-readable media are also disclosed.

    Temporal placement of a rebuffering event

    公开(公告)号:US11388472B2

    公开(公告)日:2022-07-12

    申请号:US17243389

    申请日:2021-04-28

    申请人: Netflix, Inc.

    摘要: A method includes receiving, with a computing system, data representing a video item into a buffer. The method further includes outputting the video item from the buffer to a display system. The method further includes determining that utilization of the buffer falls below a predetermined threshold. The method further includes, in response to determining that the utilization of the buffer falls below the predetermined threshold, determining that there is a specified rebuffering point within a predetermined time frame. The method further includes pausing with the computing system, the video item at the specified rebuffering point in response to determining that there is the specified rebuffering point within the predetermined time frame.

    MACHINE LEARNING TECHNIQUES FOR VIDEO DOWNSAMPLING

    公开(公告)号:US20220198607A1

    公开(公告)日:2022-06-23

    申请号:US17133206

    申请日:2020-12-23

    申请人: NETFLIX, INC.

    IPC分类号: G06T3/40 G06T9/00 G06N3/08

    摘要: In various embodiments, a training application trains a convolutional neural network to downsample images in a video encoding pipeline. The convolution neural network includes at least two residual blocks and is associated with a downsampling factor. The training application executes the convolutional neural network on a source image to generate a downsampled image. The training application then executes an upsampling algorithm on the downsampled image to generate a reconstructed image having the same resolution as the source image. The training application computes a reconstruction error based on the reconstructed image and the source image. The training application updates at least one parameter of the convolutional neural network based on the reconstruction error to generate a trained convolutional neural network. Advantageously, the trained convolution neural network can be implemented in a video encoding pipeline to mitigate visual quality reductions typically experienced with conventional video encoding pipelines that implement conventional downsampling techniques.

    TECHNIQUES FOR ADVANCING PLAYBACK OF INTERACTIVE MEDIA TITLES IN RESPONSE TO USER SELECTIONS

    公开(公告)号:US20220182737A1

    公开(公告)日:2022-06-09

    申请号:US17682237

    申请日:2022-02-28

    申请人: NETFLIX, INC.

    发明人: Mark WATSON

    摘要: A playback application seamlessly advances playback of and interactive media title in response to user selections in a manner that minimizes latency and preserves user immersion in a narrative. The playback application buffers an interstitial segment included in the interactive media title and feeds portions of the interstitial segment to a media player only when those portions are needed for display. When the user selects an option displayed during the interstitial segment, the playback application begins buffering a subsequent media segment and stops feeding portions of the interstitial segment to the media player. The playback application starts feeding blocks of the subsequent media segment to the media player and then seamlessly advances playback to the subsequent media segment.

    Audio transitions when streaming audiovisual media titles

    公开(公告)号:US11336947B2

    公开(公告)日:2022-05-17

    申请号:US16570880

    申请日:2019-09-13

    申请人: NETFLIX, INC.

    发明人: Mark Watson

    摘要: A playback application is configured to analyze audio frames associated with transitions between segments within a media title to identify one or more portions of extraneous audio. The playback application is configured to analyze the one or more portions of extraneous audio and then determine which of the one or more corresponding audio frames should be dropped. In doing so, the playback application can analyze a topology associated with the media title to determine whether any specific portions of extraneous audio are to be played outside of a logical ordering of audio samples set forth in the topology. These specific portions of extraneous audio are preferentially removed.

    Techniques for automatically extracting compelling portions of a media content item

    公开(公告)号:US11334752B2

    公开(公告)日:2022-05-17

    申请号:US16689019

    申请日:2019-11-19

    申请人: NETFLIX, INC.

    IPC分类号: G06K9/00 G06K9/62

    摘要: In various embodiments, a clip application computes a set of appearance values for an appearance metric based on shot sequences associated with a media content item. Each appearance value in the set of appearance values indicates a prevalence of a first character in a different shot sequence associated with the media content item. The clip application then performs one or more clustering operations on the shot sequences based on the set of appearance values to generate a first shot cluster. Subsequently, the clip application generates a clip for the media content item based on the first shot cluster. The clip application transmits the clip to an endpoint device for display. Advantageously, relative to primarily manual approaches, the clip application can more efficiently and reliably generate clips for media content items.