Methods and systems are described herein for hosting and arbitrating algorithms for the generation of structured frames of data from one or more sources of unstructured input frames. A plurality of frames may be received from a recording device and a plurality of object types to be recognized in the plurality of frames may be determined. A determination may be made of multiple machine learning models for recognizing the object types. The frames may be sequentially input into the machine learning models to obtain a plurality of sets of objects from the plurality of machine learning models and object indicators may be received from those machine learning models. A set of composite frames with the plurality of indicators corresponding to the plurality of objects may be generated, and an output stream may be generated including the set of composite frames to be played back in chronological order.

Aspects of the subject disclosure may include, for example, receiving a plurality of proposed machine learning solutions to a machine learning problem including receiving, for each respective proposed machine learning solution of the plurality of proposed machine learning solutions, one or more of a machine learning model, a dataset and a data pipeline output; automatically determining hybrid solutions to the machine learning problem, including combining, by the processing system, at least one of a first component from a first proposed machine learning solution with at least one of a second component from a second proposed machine learning solution; and ranking the hybrid solutions including determining a log loss score for each hybrid solution and sorting the hybrid solutions according to the log loss score for each hybrid solution. Other embodiments are disclosed.

Disclosed herein are system, method, and computer program product embodiments for a dynamic symbol-based system for objects-of-interest (OOI) video analytics detection. Some embodiments include instantiating one or more symbolic objects associated with one or more real world rules, defining an area of interest, associating one or more CV functions with the one or more symbolic objects, and identifying one or more video sources for which to apply the one or more CV functions. Some embodiments further include executing the one or more CV functions associated with the one or more symbolic objects to process the one or more video sources.

A system includes an image sensor having a plurality of pixels that form a plurality of regions of interest (ROIs), image processing resources, and a scheduler configured to perform operations including determining a priority level for a particular ROI of the plurality of ROIs based on a feature detected by one or more image processing resources of the image processing resources within initial image data associated with the particular ROI. The operations also include selecting, based on the feature detected within the initial image data, a particular image processing resource of the image processing resources by which subsequent image data generated by the particular ROI is to be processed. The operations further include inserting, based on the priority level, the subsequent image data into a processing queue of the particular image processing resource to schedule the subsequent image data for processing by the particular image processing resource.

A computer-readable recording medium having stored therein a program for causing a computer to execute a process including: receiving first image from a mobile device that photographs the first image from a variable position; transmitting the first image to a first server that executes an inference process, based on the first model, on the first image; receiving second image being same in a pixel number and a recognition target for the inference process as the first image from a fixed device that photographs the second image from a fixed position; and when determining that two of the second images received from the fixed device continuously in time series have no difference under a state where a failure of the first server is detected, transmitting the first image to a second server that executes an inference process, based on a second model, on the second image.

A scheduling algorithm for scheduling training of deep neural network (DNN) weights on processing units identifies a next job to provisionally assign a processing unit (PU) based on a doubling heuristic. The doubling heuristic makes use of an estimated number of training sets needed to complete training of weights for a given job and/or a training speed function which indicates how fast the weights are converging. The scheduling algorithm solves a problem of efficiently assigning PUs when multiple DNN weight data structures must be trained efficiently. In some embodiments, the training of the weights uses a ring-based message passing architecture. In some embodiments, performance using a nested loop approach or nested loop fashion is provided. In inner iterations of the nested loop, PUs are scheduled and jobs are launched or re-started. In outer iterations of the nested loop, jobs are stopped, parameters are updated and the inner iteration is re-entered.

A scheduling algorithm for scheduling training of deep neural network (DNN) weights on processing units identifies a next job to provisionally assign a processing unit (PU) based on a doubling heuristic. The doubling heuristic makes use of an estimated number of training sets needed to complete training of weights for a given job and/or a training speed function which indicates how fast the weights are converging. The scheduling algorithm solves a problem of efficiently assigning PUs when multiple DNN weight data structures must be trained efficiently. In some embodiments, the training of the weights uses a ring-based message passing architecture. In some embodiments, performance using a nested loop approach or nested loop fashion is provided. In inner iterations of the nested loop, PUs are scheduled and jobs are launched or re-started. In outer iterations of the nested loop, jobs are stopped, parameters are updated and the inner iteration is re-entered.

Methods, apparatus, systems and articles of manufacture are disclosed for scalable architectures for reference signature matching and updating. An example method for scalable architectures for reference signature matching and updating includes accessing site signatures to be compared to reference signatures from a first group of media sources. The example method also include determining if a first reference node is an owner of a first one of the site signatures, comparing a neighborhood of site signatures including the first site signature to reference signatures in a first subset of reference signatures when the first reference node is the owner of the first site signature, the first subset of references signatures stored in a first memory partition associated with the first reference node, and not comparing site signature to reference signatures when the first reference node is not the owner of the first one of the site signatures.

Various techniques are described for classifying synthetic data tasks and orchestrating a resource allocation between groups of eligible resources for processing the synthetic data tasks. Received synthetic data tasks can be classified by identifying a task category and a corresponding group of eligible resources (e.g., processors) for processing synthetic data tasks in the task category. For example, synthetic data tasks can include generation of source assets, ingestion of source assets, identification of variation parameters, variation of variation parameters, and creation of synthetic data. Certain categories of synthetic data tasks can be classified for processing with a particular group of eligible resources. For example, tasks to ingest synthetic data assets can be classified for processing on a CPU only, while a task to create synthetic data assets can be classified for processing on a GPU only. The synthetic data tasks can be queued and routed for processing by an eligible resource.

In a case of receiving an access request to a target image, an image access management device can provide an appropriate access right holder with an appropriate range of information by determining a browsing level with respect to the target image according to an access authority or purpose included in the access request to the target image, by an access management unit, by generating a final image that corresponds to the access authority by processing the feature vector according to the browsing level, and providing the generated final image as a response to the access request, by an image generation unit.