Abstract: Apparatus and methods for the pre-processing of image data so as to enhance quality of subsequent encoding and rendering. In one embodiment, a capture device is disclosed that includes a processing apparatus and a non-transitory computer readable apparatus comprising a storage medium have one or more instructions stored thereon. The one or more instructions, when executed by the processing apparatus, are configured to: receive captured image data (such as that sourced from two or more separate image sensors) and pre-process the data to enable stabilization of the corresponding images prior to encoding. In some implementations, the pre-processing includes combination (e.g., stitching) of the captured image data associated with the two or more sensors to facilitates the stabilization. Advantageously, undesirable artifacts such as object “jitter” can be reduced or eliminated. Methods and non-transitory computer readable apparatus are also disclosed.

Abstract: Methods and apparatus for metadata-based cinematography, production effects, shot selection, and/or other content augmentation. Effective cinematography conveys storyline, emotion, excitement, etc. Unfortunately, most amateur filmmakers lack the knowledge and ability to create cinema quality media. Various aspects of the present disclosure are directed to, among other things, rendering media based on instantaneous metadata. Unlike traditional post-processing techniques that rely on human subjectivity, some of the various techniques described herein leverage the camera's actual experiential data to enable cinema-quality post-processing for the general consuming public. Instantaneous metadata-based cinematography and shot selection advisories and architectures are also described.

Abstract: Apparatus and methods for the non-uniform downsampling of captured panoramic images. In one embodiment, a computing device is disclosed that includes a processing apparatus and a non-transitory computer readable apparatus comprising a storage medium have one or more instructions stored thereon. The one or more instructions, when executed by the processing apparatus, being configured to: receive captured images, the captured images obtained using two or more image sensors; non-uniformly downsample the received captured images; and encode the non-uniformly downsampled images. In some variants, the non-uniformly downsampled images take into account a desired area of interest within the captured images. In some implementations, the computing device includes an image capture device. Methods and non-transitory computer readable apparatus are also disclosed.

Abstract: Various examples describe interfacing at least one computing device to a plurality of human users. An interface system may receive goal data describing a shared goal shared by a plurality of users and transaction data describing a first transaction made by a first user of the plurality of users. The interface system may also receive, from a user computing device, audio data describing speech of the first user and of a second user. The interface system may extract a first word from the audio data and detect a conflict between at least two of the plurality of users based at least in part on the goal data, the transaction data, and the first word. The interface system may further determine a mediation routine for the conflict based at least in part on the goal data, the transaction data, and the first word, and execute the mediation routine.

Abstract: Methods and apparatus for image processing of spherical content via hardware acceleration components. In one embodiment, an EAC image is subdivided into facets via existing software addressing and written into the memory buffers (normally used for rectilinear cubemaps) in a graphics processing unit (GPU). The EAC facets may be translated, rotated, and/or mirrored so as to align with the expected three-dimensional (3D) coordinate space. The GPU may use existing hardware accelerator logic, parallelization, and/or addressing logic to greatly improve 3D image processing effects (such as a multi-band blend using Gaussian blurs.

Abstract: An image capture device may switch operation between a spherical capture mode or a non-spherical capture mode. Operation of the image capture device in the spherical capture mode includes generation of spherical visual content based on the visual content generated by multiple image sensors. Operation of the image capture device in the non-spherical capture mode includes generation of non-spherical visual content based on visual content generated by a single image sensor.

Abstract: Systems and techniques for data tagging and task dataset generation are described herein. A set of context elements may be obtained for a data element in a data source. The set of context elements may be evaluated using a machine learning processor to embed one or more tags into the data element. A task to be completed by the user may be identified. A set of task tags may be determined that correspond to the task. The data source may be searched using the set of task tags to select the data element. A task dataset may be generated that includes the data element in response to the search.

Abstract: Systems and methods for tagging portions a spherical video segment are provided. The spherical video segment may include visual content. User input indicating a moment in time in the spherical video segment and a viewing angle corresponding to the moment in time may be received. Tag information for the spherical video segment may be generated. The tag information may identify the moment in time and the viewing angle such that a subsequent presentation of the spherical video segment, proximate to the moment in time, includes a notification that the viewing angle is outside a display field of view based on the viewing angle being located outside the display field of view.

Abstract: Apparatus and methods for applying motion blur to overcapture content. In one embodiment, the motion blur is applied by selecting a number of frames of the captured image content for application of motion blur; selecting a plurality of pixel locations within the number of frames of the captured image content for the application of motion blur; applying motion blur to the captured image content in accordance with the selected number of frames and the selected plurality of pixel locations; and outputting the captured image content with the applied motion blur. In some implementations, motion blur is applied via implementation of a virtualized neutral density filter. Computerized devices and computer-readable apparatus for the application of motion blur are also disclosed.

Abstract: Methods and apparatus for shear correction in spherical projections. Embedded devices generally lack the compute and/or memory resources to implement two-dimensional (2D) stitches for spherical projections. Objects (such as a mounting fixture) within a certain distance of the camera may experience a 2D “tear” or “shear” artifact when stitched. Various embodiments of the present disclosure perform two orthogonal 1D stitches: (i) latitudinally across the meridian and (ii) longitudinally along the meridian to approximate the effect of a 2D stitch. Notably, the 1D stitch may be less precise than a true 2D stitch, however the image portion being stitched (e.g., the camera mount) is not the user's subject of interest and can be rendered much more loosely without adverse consequence. Temporal smoothing optimizations are additionally disclosed.

Abstract: A method may include accessing a report definition template, the report definition template identifying a set of data requirement for a report; mapping the set of data requirements to a corresponding semantic object in a semantic ontology; parsing a semantic map to determine a database table storing data for the semantic object; retrieving the data for the semantic object from the database table; generating a report data file adhering to the semantic object ontology based in part on the retrieved data; transmitting a logical location of the generated report data file, a logical location of the semantic map, and logical location of the semantic ontology to a blockchain node for adding to a report block in the blockchain.

Abstract: Apparatus and methods for the non-uniform downsampling of captured panoramic images. In one embodiment, a computing device is disclosed that includes a processing apparatus and a non-transitory computer readable apparatus comprising a storage medium have one or more instructions stored thereon. The one or more instructions, when executed by the processing apparatus, being configured to: receive captured images, the captured images obtained using two or more image sensors; non-uniformly downsample the received captured images; and encode the non-uniformly downsampled images. In some variants, the non-uniformly downsampled images take into account a desired area of interest within the captured images. In some implementations, the computing device includes an image capture device. Methods and non-transitory computer readable apparatus are also disclosed.

Abstract: A method may include receiving data from a computing device requesting approval of a loan application; accessing time-series data associated with the user from a knowledge graph; building a feature vector based on the accessed time-series data; inputting the feature vector into a machine learning model; receiving a response from the output from the machine learning mode, the output indicating a level of approval for the user with respect to the loan application; and transmitting a response to the request based on the level of approval.

Abstract: Methods and apparatus for image processing of spherical content via hardware acceleration components. In one embodiment, an EAC image is subdivided into facets via existing software addressing and written into the memory buffers (normally used for rectilinear cubemaps) in a graphics processing unit (GPU). The EAC facets may be translated, rotated, and/or mirrored so as to align with the expected three-dimensional (3D) coordinate space. The GPU may use existing hardware accelerator logic, parallelization, and/or addressing logic to greatly improve 3D image processing effects (such as a multi-band blend using Gaussian blurs.

Abstract: Methods and apparatus for shear correction in spherical projections. Embedded devices generally lack the compute and/or memory resources to implement two-dimensional (2D) stitches for spherical projections. Objects (such as a mounting fixture) within a certain distance of the camera may experience a 2D “tear” or “shear” artifact when stitched. Various embodiments of the present disclosure perform two orthogonal 1D stitches: (i) latitudinally across the meridian and (ii) longitudinally along the meridian to approximate the effect of a 2D stitch. Notably, the 1D stitch may be less precise than a true 2D stitch, however the image portion being stitched (e.g., the camera mount) is not the user's subject of interest and can be rendered much more loosely without adverse consequence. Temporal smoothing optimizations are additionally disclosed.

Abstract: Apparatus and methods for enabling indexing and playback of media content before the end of a content capture. In one aspect, a method for enabling indexing of media data obtained as part of a content capture is disclosed. In one embodiment, the indexing enables playback of the media data during the capture and before cessation thereof. In one variant, the method includes generating an “SOS track” for one or more images. The SOS track does not contain the same information as a full index, but provides sufficient information to allow an index to be subsequently constructed. In one implementation, the provided information includes identifiable markers relating to video data, audio data, or white space, but it does not provide an enumerated or complete “table of contents” as in a traditional index.

Abstract: Systems and methods are provided that capture and process frames of frame data. An image sensor captures frames of frame data representative of light incident upon the image sensor using a rolling shutter and outputs the frames of frame data. The image sensor captures at least one of the frames over a frame capture interval and then waits over a blanking interval before capturing another frame. A buffer receives and stores the frames output by the image sensor. An image signal processor retrieves the frames from the buffer and processes the frames over successive frame processing intervals to generate a video having a time interval per frame greater than the frame capture interval. At least one of the successive frame processing intervals is greater than the frame capture interval and is less than or equal to a sum of the frame capture interval and the blanking interval.

Abstract: An image capture device may receive GPS data during capture of video frames. GPS data may be used to increase accuracy of GPS time of the image capture device. GPS time of the image capture device at a later moment in the capture duration may be used to determine the times of earlier captured video frames.

Abstract: An image capture device may capture visual content through a front-facing optical element. A portion of the visual content may be enlarged for presenting on a front-facing display of the image capture device. Extent of the visual content within the portion may be warped to increases size of depiction within the portion.

Abstract: An image capture device may switch operation between a spherical capture mode or a non-spherical capture mode. Operation of the image capture device in the spherical capture mode includes generation of spherical visual content based on the visual content generated by multiple image sensors. Operation of the image capture device in the non-spherical capture mode includes generation of non-spherical visual content based on visual content generated by a single image sensor.