Abstract: A media item to be provided to users of a platform is identified. The media item is associated with a media class of one or more media classes. An indication of the media item is provided as input to a machine learning model trained based on historical encoding data to predict, for a given media item, a set of encoder parameter settings that satisfy a performance criterion in view of a respective media class of the given media item. The historical encoding data includes a prior set of encoder parameter settings that satisfied the performance criterion with respect to a prior media item associated with the respective class. Encoder parameter settings that satisfy the performance criterion in view of the media class is determined based on an output of the model. The media item is caused to be encoded using the determined encoder parameter settings.

Abstract: Methods and systems for encoder parameter setting optimization. A media item to be provided to one or more users of a platform is identified. The media item is associated with a media class. An indication of the identified media item is provided as input to a first machine learning model. The first machine learning model is trained to predict, for a given media item, a set of encoder parameter settings that satisfy a performance criterion in view of a respective media class associated with the given media item. One or more outputs of the first machine learning model are obtained. The one or more obtained outputs include encoder data identifying one or more sets of encoder parameter settings and, for each of the sets of encoder parameter settings, an indication of a level of confidence that a respective set of encoder parameter settings satisfies the performance criterion in view of the media class associated with the identified media item.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for encoding video comprising a sequence of video frames. In one aspect, a method comprises for one or more of the video frames: obtaining a feature embedding for the video frame; processing the feature embedding using a rate control machine learning model to generate a respective score for each of multiple quantization parameter values; selecting a quantization parameter value using the scores; determining a cumulative amount of data required to represent: (i) an encoded representation of the video frame and (ii) encoded representations of each preceding video frame; determining, based on the cumulative amount of data, that a feedback control criterion for the video frame is satisfied; updating the selected quantization parameter value; and processing the video frame using an encoding model to generate the encoded representation of the video frame.

Abstract: Methods and systems for encoder parameter setting optimization. A media item to be provided to one or more users of a platform is identified. The media item is associated with a media class. An indication of the identified media item is provided as input to a first machine learning model. The first machine learning model is trained to predict, for a given media item, a set of encoder parameter settings that satisfy a performance criterion in view of a respective media class associated with the given media item. One or more outputs of the first machine learning model are obtained. The one or more obtained outputs include encoder data identifying one or more sets of encoder parameter settings and, for each of the sets of encoder parameter settings, an indication of a level of confidence that a respective set of encoder parameter settings satisfies the performance criterion in view of the media class associated with the identified media item.

Abstract: A first aspect is a method for coding a group of pictures (GOP) that includes frames of a video. The method includes encoding, at least some of the frames of the GOP, using a first encoding pass to obtain encoding statistics; obtaining, using the encoding statistics, respective temporal dependency likelihoods (TDLs) for the at least some of the frames of the GOP, where the respective TDLs indicate contributions that the at least some of the frames make in reducing prediction errors of the GOP; obtaining a reference frame based on the respective TDLs; and using the reference frame in encoding at least some of the frames of the GOP in a second encoding pass.

Abstract: A first aspect is a method for coding a group of pictures (GOP) that includes frames of a video. The method includes encoding, at least some of the frames of the GOP, using a first encoding pass to obtain encoding statistics; obtaining, using the encoding statistics, respective temporal dependency likelihoods (TDLs) for the at least some of the frames of the GOP, where the respective TDLs indicate contributions that the at least some of the frames make in reducing prediction errors of the GOP; obtaining a reference frame based on the respective TDLs; and using the reference frame in encoding at least some of the frames of the GOP in a second encoding pass.

Abstract: A video stream of a scene for a virtual reality or augmented reality experience may be captured by one or more image capture devices. Data from the video stream may be retrieved, including base vantage data with base vantage color data depicting the scene from a base vantage location, and target vantage data with target vantage color data depicting the scene from a target vantage location. The base vantage data may be reprojected to the target vantage location to obtain reprojected target vantage data. The reprojected target vantage data may be compared with the target vantage data to obtain residual data. The residual data may be compressed by removing a subset of the residual data that is likely to be less viewer-discernable than a remainder of the residual data. A compressed video stream may be stored, including the base vantage data and the compressed residual data.

Abstract: A virtual reality or augmented reality experience of a scene may be presented to a viewer using layered data retrieval and/or processing. A first layer of a video stream may be retrieved, and a first viewer position and/or orientation may be received. The first layer may be processed to generate first viewpoint video of the scene from a first virtual viewpoint corresponding to the first viewer position and/or orientation. The first viewpoint video may be displayed for the viewer. Then, a second layer of the video stream may be retrieved, and a second viewer position and/or orientation may be received. The second layer may be processed to generate second viewpoint video of the scene from a second virtual viewpoint corresponding to the second viewer position and/or orientation, with higher quality than the first viewpoint video. The second viewpoint video may be displayed for the viewer.

Abstract: A virtual reality or augmented reality experience of a scene may be decoded for playback for a viewer through a combination of CPU and GPU processing. A video stream may be retrieved from a data store. A first viewer position and/or orientation may be received from an input device, such as the sensor package on a head-mounted display (HMD). At a processor, the video stream may be partially decoded to generate a partially-decoded bitstream. At a graphics processor, the partially-decoded bitstream may be further decoded to generate viewpoint video of the scene from a first virtual viewpoint corresponding to the first viewer position and/or orientation. The viewpoint video may be displayed on a display device, such as screen of the HMD.

Abstract: A combined video of a scene may be generated for applications such as virtual reality or augmented reality. In one method, a data store may store video data with a first portion having a first importance metric, and a second portion having a second importance metric, denoting that viewing of the first portion is more likely and/or preferential to viewing of the second portion. The subset may be retrieved and used to generate viewpoint video from a virtual viewpoint corresponding to a viewer's viewpoint. The viewpoint video may be displayed on a display device. One of storing the video data, retrieving the subset, and using the subset to generate the viewpoint video may include, based on the difference between the first and second importance metrics, expediting and/or enhancing performance of the step for the first portion, relative to the second portion.

Abstract: A video stream for a scene for a virtual reality or augmented reality experience may be stored and delivered to a viewer. The video stream may be divided into a plurality of units based on time segmentation, viewpoint segmentation, and/or view orientation segmentation. Each of the units may be divided into a plurality of sub-units based on a different segmentation from the units, via time segmentation, viewpoint segmentation, and/or view orientation segmentation. At least a portion of the video stream may be stored in a file that includes a plurality of the units. Each unit may be a group of pictures that is a sequence of successive frames in time. Each sub-unit may be a vantage defining a viewpoint from which the scene is viewable. Each vantage may be further divided into tiles, each of which is part of the vantage, limited to one or more particular view orientations.

Abstract: An environment may be displayed from a viewpoint. According to one method, volumetric video data may be acquired depicting the environment, for example, using a tiled camera array. A plurality of vantages may be distributed throughout a viewing volume from which the environment is to be viewed. The volumetric video data may be used to generate video data for each vantage, representing the view of the environment from that vantage. User input may be received designating a viewpoint within the viewing volume. From among the plurality of vantages, a subset nearest to the viewpoint may be identified. The video data from the subset may be retrieved and combined to generate viewpoint video data depicting the environment from the viewpoint. The viewpoint video data may be displayed for the viewer to display a view of the environment from the viewpoint selected by the user.

Abstract: A combined video of a scene may be generated for applications such as virtual reality or augmented reality. In one method, a data store may store video data with a first portion having a first importance metric, and a second portion having a second importance metric, denoting that viewing of the first portion is more likely and/or preferential to viewing of the second portion. The subset may be retrieved and used to generate viewpoint video from a virtual viewpoint corresponding to a viewer's viewpoint. The viewpoint video may be displayed on a display device. One of storing the video data, retrieving the subset, and using the subset to generate the viewpoint video may include, based on the difference between the first and second importance metrics, expediting and/or enhancing performance of the step for the first portion, relative to the second portion.

Abstract: A video stream of a scene for a virtual reality or augmented reality experience may be captured by one or more image capture devices. Data from the video stream may be retrieved, including base vantage data with base vantage color data depicting the scene from a base vantage location, and target vantage data with target vantage color data depicting the scene from a target vantage location. The base vantage data may be reprojected to the target vantage location to obtain reprojected target vantage data. The reprojected target vantage data may be compared with the target vantage data to obtain residual data. The residual data may be compressed by removing a subset of the residual data that is likely to be less viewer-discernable than a remainder of the residual data. A compressed video stream may be stored, including the base vantage data and the compressed residual data.

Abstract: A virtual reality or augmented reality experience of a scene may be presented to a viewer using layered data retrieval and/or processing. A first layer of a video stream may be retrieved, and a first viewer position and/or orientation may be received. The first layer may be processed to generate first viewpoint video of the scene from a first virtual viewpoint corresponding to the first viewer position and/or orientation. The first viewpoint video may be displayed for the viewer. Then, a second layer of the video stream may be retrieved, and a second viewer position and/or orientation may be received. The second layer may be processed to generate second viewpoint video of the scene from a second virtual viewpoint corresponding to the second viewer position and/or orientation, with higher quality than the first viewpoint video. The second viewpoint video may be displayed for the viewer.

Abstract: A virtual reality or augmented reality experience of a scene may be decoded for playback for a viewer through a combination of CPU and GPU processing. A video stream may be retrieved from a data store. A first viewer position and/or orientation may be received from an input device, such as the sensor package on a head-mounted display (HMD). At a processor, the video stream may be partially decoded to generate a partially-decoded bitstream. At a graphics processor, the partially-decoded bitstream may be further decoded to generate viewpoint video of the scene from a first virtual viewpoint corresponding to the first viewer position and/or orientation. The viewpoint video may be displayed on a display device, such as screen of the HMD.

Abstract: A video stream for a scene for a virtual reality or augmented reality experience may be stored and delivered to a viewer. The video stream may be divided into a plurality of units based on time segmentation, viewpoint segmentation, and/or view orientation segmentation. Each of the units may be divided into a plurality of sub-units based on a different segmentation from the units, via time segmentation, viewpoint segmentation, and/or view orientation segmentation. At least a portion of the video stream may be stored in a file that includes a plurality of the units. Each unit may be a group of pictures that is a sequence of successive frames in time. Each sub-unit may be a vantage defining a viewpoint from which the scene is viewable. Each vantage may be further divided into tiles, each of which is part of the vantage, limited to one or more particular view orientations.

Abstract: An environment may be displayed from a viewpoint. According to one method, volumetric video data may be acquired depicting the environment, for example, using a tiled camera array. A plurality of vantages may be distributed throughout a viewing volume from which the environment is to be viewed. The volumetric video data may be used to generate video data for each vantage, representing the view of the environment from that vantage. User input may be received designating a viewpoint within the viewing volume. From among the plurality of vantages, a subset nearest to the viewpoint may be identified. The video data from the subset may be retrieved and combined to generate viewpoint video data depicting the environment from the viewpoint. The viewpoint video data may be displayed for the viewer to display a view of the environment from the viewpoint selected by the user.

Abstract: A method and apparatus are provided that are implemented at a conference apparatus to provide an optimization of video conferencing resource utilization. The method receives a plurality of resource constraints where each resource constraint is from a conference device in a plurality of conference devices, wherein at least a first resource constraint of a first conference device contains an indication of a video processing capability of the first conference device. The method further computes a conference solution matrix for the plurality of conference devices based on the plurality of resource constraints, wherein the conference solution matrix contains a solution entry for each of the plurality of conference devices, and wherein for the first conference device, a corresponding first solution entry indicates a video processing solution selection for the first conference device. Solution entries are then sent to corresponding conference devices of the plurality of conference devices.

Abstract: A method of proximity-based conference session transfer is disclosed. The method is implemented at a conference device and it starts with the conference device generating a conference session code based on an encrypted conference session message received from a conference server. The conference device communicates the conference session code to an area in proximity to the conference device, where a user device in the area in proximity to the conference device receives the conference session code and requests the conference server to coordinate the conference device to assume a conference session of the user device. The conference device then receives a message from the conference server requesting the conference device to assume the conference session of the user device, and it assumes the conference session of the user device by performing conference functions according to input from the user device.