Abstract: Techniques for migrating live broadcast streams from one data center to another data center are described herein. A data center from a plurality of data centers may be determined to transfer live streaming video content from a first data center to the data center. A separate live streaming session for each live streaming session being broadcast by the first data center may be generated by communicating with an ingest service server that receives media content used to generate each separate live streaming session and the live streaming session. Each separate live streaming session may be routed to the candidate data center. A determination that each separate live streaming session of the candidate data center is synchronized with each live streaming session of the first data center may be made. An update to a mapping of IP addresses to redirect requests from the first data center to the candidate data center.

Abstract: A video streaming service may include a segmentation component that provides segment metadata that indicates starting locations for a plurality of segments of video content that are produced during transcoding of the video content. The segment metadata may be sent to a first transcode server and a second transcode server to allow alignment of segmentations between the first transcode server and the second transcode server. This may allow transcoding of the video content to be seamlessly switched from the first transcode server to the second transcode server, during transmission of the video content, without causing interruption of playback. The first transcode server and the second transcode server may also operate in parallel to produce different versions of the video content, such as versions encoded using different codecs. A video player may seamlessly switch between these different codec versions without causing interruption of playback.

Abstract: The disclosed computer-implemented method may include accessing media segments that correspond to respective media items. At least one of the media segments may be divided into discrete video shots. The method may also include matching the discrete video shots in the media segments to corresponding video shots in the corresponding media items according to various matching factors. The method may further include generating a relative similarity score between the matched video shots in the media segments and the corresponding video shots in the media items, and training a machine learning model to automatically identify video shots in the media items according to the generated relative similarity score between matched video shots. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented method may include accessing media segments that correspond to respective media items. At least one of the media segments may be divided into discrete video shots. The method may also include matching the discrete video shots in the media segments to corresponding video shots in the corresponding media items according to various matching factors. The method may further include generating a relative similarity score between the matched video shots in the media segments and the corresponding video shots in the media items, and training a machine learning model to automatically identify video shots in the media items according to the generated relative similarity score between matched video shots. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented method may include generating a three-dimensional (3D) feature map for a digital image using a fully convolutional network (FCN). The 3D feature map may be configured to identify features of the digital image and identify an image region for each identified feature. The method may also include generating a region composition graph that includes the identified features and image regions. The region composition graph may be configured to model mutual dependencies between features of the 3D feature map. The method may further include performing a graph convolution on the region composition graph to determine a feature aesthetic value for each node according to the weightings in the node's weighted connecting segments, and calculating a weighted average for each node's feature aesthetic value to provide a combined level of aesthetic appeal for the digital image. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A content production system and content management system are provided to produce and manage content with multiple components, such as multi-track content with any number or combination of video tracks, audio tracks, and the like. Individual tracks of the content are selectively included in, or excluded from, different versions of the content such that different versions of the multi-track content may be generated from different subsets of the individual tracks. For example, one version of content may be generated for real-time streaming and may include a first subset or combination of the multiple tracks. Another version of the content may be generated for storage and future on-demand access, and may include a second subset or combination of the multiple tracks. Selection of the particular tracks of the multi-track content to be included in each version may be based on varying types of authorization associated with the individual tracks.

Abstract: The disclosed computer-implemented method may include accessing media segments that correspond to respective media items. At least one of the media segments may be divided into discrete video shots. The method may also include matching the discrete video shots in the media segments to corresponding video shots in the corresponding media items according to various matching factors. The method may further include generating a relative similarity score between the matched video shots in the media segments and the corresponding video shots in the media items, and training a machine learning model to automatically identify video shots in the media items according to the generated relative similarity score between matched video shots. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: The disclosed computer-implemented method may include accessing media segments that correspond to respective media items. At least one of the media segments may be divided into discrete video shots. The method may also include matching the discrete video shots in the media segments to corresponding video shots in the corresponding media items according to various matching factors. The method may further include generating a relative similarity score between the matched video shots in the media segments and the corresponding video shots in the media items, and training a machine learning model to automatically identify video shots in the media items according to the generated relative similarity score between matched video shots. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a defective pixel detection application automatically detects defective pixels in video content. In operation, the defective pixel detection application computes a first set of pixel intensity gradients based on a first frame of video content and a first neighborhood of pixels associated with a first pixel. The defective pixel detection application also computes a second set of pixel intensity gradients based on the first frame and a second neighborhood of pixels associated with the first pixel. Subsequently, the defective pixel detection application computes a statistical distance between the first set of pixel intensity gradients and the second set of pixel intensity gradients. The defective pixel detection application then determines that the first pixel is defective based on the statistical distance.

Abstract: The disclosed computer-implemented method may include accessing media segments that correspond to respective media items. At least one of the media segments may be divided into discrete video shots. The method may also include matching the discrete video shots in the media segments to corresponding video shots in the corresponding media items according to various matching factors. The method may further include generating a relative similarity score between the matched video shots in the media segments and the corresponding video shots in the media items, and training a machine learning model to automatically identify video shots in the media items according to the generated relative similarity score between matched video shots. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: One embodiment of the present invention sets forth a technique for detecting a shot change in a video. The technique includes calculating a first plurality of difference values for pixel blocks included in a first video frame and a second video frame and calculating a first cumulative distribution function (CDF) based on the first plurality of difference values. The technique further includes calculating a second plurality of difference values for pixel blocks included in a first plurality of video frames that is contiguous with the first video frame and calculating a second CDF based on the second plurality of difference values. The technique further includes comparing the first CDF to the second CDF to detect a shot change.

Abstract: A neural network system that is trained to identify one or more portions of a media title where synchronization errors are likely to be present. The neural network system is trained based on a first set of media titles where synchronization errors are present and a second set of media titles where synchronization errors are absent. The second set of media titles can be generated by introducing synchronization errors into a set of media titles that otherwise lack synchronization errors. Via training, the neural network system learns to identify specific visual features included in one or more video frames and corresponding audio features that should be played back in synchrony with the associated visual features. Accordingly, when presented with a media title that includes synchronization errors, the neural network can indicate the specific frames where synchronization errors are likely to be present.

Abstract: The disclosed computer-implemented method may include generating a three-dimensional (3D) feature map for a digital image using a fully convolutional network (FCN). The 3D feature map may be configured to identify features of the digital image and identify an image region for each identified feature. The method may also include generating a region composition graph that includes the identified features and image regions. The region composition graph may be configured to model mutual dependencies between features of the 3D feature map. The method may further include performing a graph convolution on the region composition graph to determine a feature aesthetic value for each node according to the weightings in the node's weighted connecting segments, and calculating a weighted average for each node's feature aesthetic value to provide a combined level of aesthetic appeal for the digital image. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: Systems, methods and computer-readable storage devices each provide, for a given flow entry in a flow table associated with a packet flow, a first atomic counter associated with a first color and a second atomic counter associated with a second color. The system, for a first coloring interval, marks traffic in the packet flow from a tenant to a fabric as the first color and increments the first atomic counter. At a conclusion of the first coloring interval and for a second coloring interval, the system marks the traffic in the packet flow from the tenant to the fabric as the second color and increments the second atomic counter. The system compares first packet counts associated with the first color at an ingress point with second packet counts associated with the first color at an egress point to the fabric.

Abstract: In various embodiments, a defective pixel detection application automatically detects defective pixels in video content. In operation, the defective pixel detection application computes a first set of pixel intensity gradients based on a first frame of video content and a first neighborhood of pixels associated with a first pixel. The defective pixel detection application also computes a second set of pixel intensity gradients based on the first frame and a second neighborhood of pixels associated with the first pixel. Subsequently, the defective pixel detection application computes a statistical distance between the first set of pixel intensity gradients and the second set of pixel intensity gradients. The defective pixel detection application then determines that the first pixel is defective based on the statistical distance.

Abstract: One embodiment of the present invention sets forth a technique for mitigating drift in audiovisual assets. The technique includes determining that an edit associated with a presentation timeline is within boundaries of a video frame. The technique further includes calculating a temporal drift associated with the edit, where the temporal drift comprises a duration of time between the edit and a boundary of the video frame. The technique further includes determining whether to include the video frame in the presentation timeline based on the temporal drift and an accumulated temporal drift associated with the presentation timeline.

Abstract: One embodiment of the present invention sets forth a technique for detecting a shot change in a video. The technique includes calculating a first plurality of difference values for pixel blocks included in a first video frame and a second video frame and calculating a first cumulative distribution function (CDF) based on the first plurality of difference values. The technique further includes calculating a second plurality of difference values for pixel blocks included in a first plurality of video frames that is contiguous with the first video frame and calculating a second CDF based on the second plurality of difference values. The technique further includes comparing the first CDF to the second CDF to detect a shot change.

Abstract: Systems, methods and computer-readable storage devices each provide, for a given flow entry in a flow table associated with a packet flow, a first atomic counter associated with a first color and a second atomic counter associated with a second color. The system, for a first coloring interval, marks traffic in the packet flow from a tenant to a fabric as the first color and increments the first atomic counter. At a conclusion of the first coloring interval and for a second coloring interval, the system marks the traffic in the packet flow from the tenant to the fabric as the second color and increments the second atomic counter. The system compares first packet counts associated with the first color at an ingress point with second packet counts associated with the first color at an egress point to the fabric.

Abstract: One embodiment of the present invention sets forth a technique for mitigating drift in audiovisual assets. The technique includes determining that an edit associated with a presentation timeline is within boundaries of a video frame. The technique further includes calculating a temporal drift associated with the edit, where the temporal drift comprises a duration of time between the edit and a boundary of the video frame. The technique further includes determining whether to include the video frame in the presentation timeline based on the temporal drift and an accumulated temporal drift associated with the presentation timeline.

Abstract: In one embodiment, a method includes receiving at a multimedia transformation unit, media streams from a plurality of endpoints, transmitting audio components of the media streams to a multipoint conferencing unit, receiving an identifier from the multipoint conferencing unit identifying one of the media streams as an active speaker stream, processing at the multimedia transformation unit, a video component of the active speaker stream, and transmitting the active speaker stream to one or more of the endpoints without transmitting the video component to the multipoint conferencing unit. An apparatus is also disclosed herein.