Abstract: Techniques are disclosed for multiplexing a dynamic bit-rate video stream with an audio stream received by a client device in a manner that allows the resulting multiplexed stream to be played back without disruption, despite dynamic changes in the bit rate of the video stream that may occur. A content server may stream both a video stream and an audio stream to a client device for playback. The client device may multiplex the video and audio streams prior to them being presented to a playback engine for decoding and playback to a user.

Abstract: In various embodiments, a network emulation application emulates network conditions when testing a software application. In response to a request to emulate a first set of network conditions for a first client device that is executing the software application, causing a kernel to implement a first pipeline and to automatically input network traffic associated with the first client device to the first pipeline instead of a default bridge. In response to a request to emulate a second set of network conditions for a second client device that is executing the software application, causing the kernel to implement a second pipeline and to automatically input network traffic associated with the second client device to the second pipeline instead of the default bridge. Each of the pipelines perform one or more traffic shaping operations on at least a subset of the network traffic input into the pipeline.

Abstract: The disclosed computer-implemented method may include detecting an application running in a background state on a client device. The method may also include collecting state data about a current state of the client device. Additionally, the method may include determining, by applying a machine learning model to the collected state data, that a likelihood of forcible termination of the application within a predetermined timeframe exceeds a threshold. Furthermore, the method may include reducing a computing resource footprint of the application on the client device to reduce the likelihood of forcible termination of the application. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a training application trains a machine learning model to preprocess images. In operation, the training application computes a chroma sampling factor based on a downscaling factor and a chroma subsampling ratio. The training application executes a machine learning model that is associated with the chroma sampling factor on data that corresponds to both an image and a first chroma component to generate preprocessed data corresponding to the first chroma component. Based on the preprocessed data, the training application updates at least one parameter of the machine learning model to generate a trained machine learning model that is associated with the first chroma component.

Abstract: In various embodiments, a training application generates a perceptual video model. The training application computes a first feature value for a first feature included in a feature vector based on a first color component associated with a first reconstructed training video. The training application also computes a second feature value for a second feature included in the feature vector based on a first brightness component associated with the first reconstructed training video. Subsequently, the training application performs one or more machine learning operations based on the first feature value, the second feature value, and a first subjective quality score for the first reconstructed training video to generate a trained perceptual quality model. The trained perceptual quality model maps a feature value vector for the feature vector to a perceptual quality score.

Abstract: In various embodiments, a training application generates a preference prediction model based on an interaction matrix and a closed-form solution for minimizing a Lagrangian. The interaction matrix reflects interactions between users and items, and the Lagrangian is formed based on a constrained optimization problem associated with the interaction matrix. A service application generates a first application interface that is to be presented to the user. The service application computes predicted score(s) using the preference prediction model, where each predicted score predicts a preference of the user for a different item. The service application then determines a first item from the items to present to the user via an interface element included in the application interface. Subsequently, the service application causes a representation of the first item to be displayed via the interface element included in the application interface.

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: In various embodiments, a shot collation application causes multiple encoding instances to encode a source video sequence that includes at least two shot sequences. The shot collation application assigns a first shot sequence to a first chunk. Subsequently, the shot collation application determines that a second shot sequence does not meet a collation criterion with respect to the first chunk. Consequently, the shot collation application assigns the second shot sequence or a third shot sequence derived from the second shot sequence to a second chunk. The shot collation application causes a first encoding instance to independently encode each shot sequence assigned to the first chunk. Similarly, the shot collation application causes a second encoding instance to independently encode each shot sequence assigned to the second chunk. Finally, a chunk assembler combines the first encoded chunk and the second encoded chunk to generate an encoded video sequence.

Abstract: In various embodiments, a quality inference application estimates the perceived quality of reconstructed videos. The quality inference application computes a first feature value for a first feature included in a feature vector based on a color component associated with a reconstructed video. The quality inference application also computes a second feature value for a second feature included in the feature vector based on a brightness component associated with the reconstructed video. Subsequently, the quality inference application computes a perceptual quality score based on the first feature value and the second feature value. The perceptual quality score indicates a level of visual quality associated with at least one frame included in the reconstructed video.

Abstract: One embodiment of the present invention includes a server machine configured to establish a secure communication channel with a client machine via renewable tokens. The server machine receives a plurality of messages from a client machine over a secure communication channel, where the plurality of messages includes a first message that includes at least two of user authentication data, entity authentication data, first key exchange data, and encrypted message data. The server machine transmits, to the client machine, a second message that includes a master token comprising second key exchange data associated with the first key exchange data and at least one of a renewal time and an expiration time.

Abstract: The disclosed computer-implemented method includes determining that audio quality is to be adjusted for a multimedia streaming connection over which audio data and video data are being streamed to a content player. The audio data is streamed at a specified audio quality level and the video data is streamed at a specified video quality level. The method also includes determining that a specified minimum video quality level is to be maintained while adjusting the audio quality level. Still further, the method includes dynamically adjusting the audio quality level of the multimedia streaming connection while maintaining the video quality level of the multimedia streaming connection at at least the specified minimum video quality level. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a hindsight application computes a total download size for a sequence of encoded chunks associated with a media title for evaluation of at least one aspect of a video streaming service. The hindsight application computes a feasible download end time associated with a source chunk of the media title based on a network throughput trace and a subsequent feasible download end time associated with a subsequent source chunk of the media title. The hindsight application then selects an encoded chunk associated with the source chunk based on the network throughput trace, the feasible download end time, and a preceding download end time associated with a preceding source chunk of the media title. Subsequently, the hindsight application computes the total download size based on the number of encoded bits included in the first encoded chunk. The total download size correlates to an upper bound on visual quality.

Abstract: A system for utilizing models derived from offline historical data in online applications is provided. The system includes a processor and a memory storing machine-readable instructions for determining a set of contexts of the usage data, and for each of the contexts within the set of contexts, collecting service data from services supporting the media service and storing that service data in a database. The system performing an offline testing process by fetching service data for a defined context from the database, generating a first set of feature vectors based on the fetched service data, and providing the first set to a machine-learning module. The system performs an online testing process by fetching active service data from the services supporting the media streaming service, generating a second set of feature vectors based on the fetched active service data, and providing the second set to the machine-learning module.

Abstract: Techniques are described for detecting faults in media content based on the behavior of users viewing the media content. Embodiments stream a first instance of media content to one or more users. The behavior of the one or more users is monitored while the users are viewing the streaming first instance of media content. Embodiments then determine whether the first instance of media content is faulty, based on the monitored behavior of the one or more users.

Abstract: In various embodiments, a perceptual quality application computes an absolute quality score for encoded video content. In operation, the perceptual quality application selects a model based on the spatial resolution of the video content from which the encoded video content is derived. The model associates a set of objective values for a set of objective quality metrics with an absolute quality score. The perceptual quality application determines a set of target objective values for the objective quality metrics based on the encoded video content. Subsequently, the perceptual quality application computes the absolute quality score for the encoded video content based on the selected model and the set of target objective values. Because the absolute quality score is independent of the quality of the video content, the absolute quality score accurately reflects the perceived quality of a wide range of encoded video content when decoded and viewed.

Abstract: 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.

Abstract: In various embodiments, a forensic scoping application analyzes host instances in order to detect anomalies. The forensic scoping application acquires a snapshot for each host instance included in an instance group. Each snapshot represents a current operational state of the associated host instance. Subsequently, the forensic scoping application performs clustering operation(s) based on the snapshots to generate a set of clusters. The forensic scoping application determines that a first cluster in the set of clusters is associated with fewer host instances than at least a second cluster in the set of clusters. Based on the first cluster, the forensic scoping application determines that a first host instance included in the instance group is operating in an anomalous fashion. Advantageously, efficiently determining host instances that are operating in an anomalous fashion during a security attack can reduce the amount of damage caused by the security attack.

Abstract: In various embodiments, a codec comparison application independently encodes each source image included in a set of source images using a first encoding configuration to generate a first set of encoded images. The codec comparison application also independently encodes each source image included in the set of source images using a second encoding configuration to generate a second set of encoded images. For each encoded image in the first set of encoded images and each encoded image in the second set of encoded images, a visual quality score for a reconstructed source image derived from the encoded image falls within a tolerance of a target visual quality score.

Abstract: The disclosed computer-implemented method may include receiving, as an input, segmented video scenes, where each video scene includes a specified length of video content. The method may further include scanning the video scenes to identify objects within the video scene and also determining a relative importance value for the identified objects. The relative importance value may include an indication of which objects are to be included in a cropped version of the video scene. The method may also include generating a video crop that is to be applied to the video scene such that the resulting cropped version of the video scene includes those identified objects that are to be included based on the relative importance value. The method may also include applying the generated video crop to the video scene to produce the cropped version of the video scene. Various other methods, systems, and computer-readable media are also disclosed.