Abstract: Presented herein are systems and methods for receiving video data associated with at least one task performed by at least one worker and executing, based on the received video data, a human action recognition program to identify at least one action associated with the at least one task and executing an object detection program to identify at least one object associated with the at least one task, identifying, based on a combination of the identified at least one action and the identified at least one object associated with the at least one task, at least one subtask of the at least one task, and generating at least first and second labels for the identified at least one subtask based on at least first and second labels, respectively, associated from a first set of existing labels for identified actions and from a second set of existing labels for identified objects, respectively.

Abstract: A method for object defect detection. The method may include receiving an object on a production line; computing, by a processor, a motion optimized path for a robot arm, wherein the motion optimized path comprises a path for performing a sequence of rotations by the robot arm on the object for image capturing; using the robot arm to grasp the object and moving the robot arm according to the motion optimized path to rotate the object based on the sequence of rotations; capturing, by a camera, a plurality of images of the object while the object is being rotated; performing, by the processor, defect detection on the plurality of images of the object to determine object defect; and for object defect being detected, issuing, by the processor, a defect notification to an operator of the production line.

Abstract: Systems and methods for automating process setting to a target factory, which can involve creating templatized business terms, templatized business data configurator logics, and a templatized data profile by machine learning from training data from at least one reference factory; storing the templatized business terms, the templatized business data configurator logics, and the templatized data profile into a knowledge graph; querying the knowledge graph with a data profile of the target factory to obtain corresponding templated business terms; and applying the corresponding templated business terms and corresponding templated business data configurator logics to a data catalogue of the target factory.

Abstract: A method for generating query response comprising receiving, by a processor, an input query from a first user; performing, by the processor, query parsing on the input query to generate a parsed query; determining, by the processor, a query type associated with the parsed query; for the query type being determined as initial query, performing: generating, by the processor, a first follow-up query to the input query based on the parsed query, and generating, by the processor, responses to the parsed query and the first follow-up query; and for the query type being determined as follow-up query to a previous query entered by a second user, performing: performing, by the processor, learning of a query sequence from the previous query to the input query, and generating, by the processor, responses to the parsed query and a second follow-up query.

Abstract: Example implementations described herein involve systems and methods to substantially simultaneously orchestrate machine learning models over multiple resource constrained control edge devices, so that the overall system is more agile to changes in events and environmental conditions where the models have been deployed. The example implementations described herein involve multiple processes that when executed, determine a list of edge devices to be updated along with the corresponding models based on correlation.

Abstract: A method for computing and detecting image data drift. The method may include retrieving first segment information of a plurality of segments from a drift database; receiving a number of images from a sensor; partitioning each of the received images into segments of a predetermined number; generating second segment information; computing drift in values between the first segment information and the second segment information; and detecting drift based on the computed drift in values by combining the computed drift in segments to generate overall drift, and comparing the overall drift against a drift threshold.

Abstract: A method for generating and displaying context-based information. The method may include automatically identifying, by a processor, a person; identifying, by the processor, a display device closest in proximity to the person; generating, by the processor, context information specific to the person; and displaying the context information on the display device to be viewed by the person.

Abstract: Example implementations described herein involve systems and methods to select machine learning models that will be executed in a cellular Mobile Edge Computing for cellular enabled applications. In contrast to related art implementations, the example implementations described herein considers different data fidelities of received data due to the cellular wireless channel and also performs service resource allocation accordingly.

Abstract: In example implementations described herein, there are systems and methods for generating at least a first set of weights for a first neural network associated with a first task performed in a first environment and a second set of weights for a second neural network associated with the first task performed in a second environment; training a metamodel based on at least the first set of weights and the second set of weights; and generating, based on the metamodel, a third set of weights for a third neural network associated with a second task in the second environment.

Abstract: Example implementations described herein involve systems and methods that involve recognizing, from sensor data, an area from the plurality of areas and a candidate task from the one or more candidate tasks associated with the area; estimating a probability of each of the plurality of candidate tasks for the each of the plurality of areas for a specific future period of time, based on referencing historical data of task sequences previously executed; accepting the ones of the plurality of candidate tasks for the each of the plurality of areas having the probability being higher than a threshold; and scheduling one or more sensors to activate and transmit in the specific future period of time in associated areas for the plurality of areas associated with other ones of the plurality of candidate tasks for the each of the plurality of areas not having the probability being higher than the threshold.

Abstract: Example implementations described herein involve systems and methods for providing a reward to a machine learning algorithm, which can include receiving an image, and a task description defined in text; slicing the image into a plurality of sub-images; executing an embedding model to embed the text of the task description and the sub-images to generate a distribution for the sub-images based on relevance to the task description; and generating the reward from the distribution for the sub-images.

Abstract: Example implementations described herein can dynamically adapt to changing nature of sensor data traffic and through artificial intelligence (AI, strike a good tradeoff between reducing volume of sensed data, and retain enough data fidelity so that subsequent analytics applications perform well. The example implementations eliminate heuristic methods of setting sensing parameters (such as DAQ sampling rate, resolution etc.) and replaces them with an automated, AI driven edge solution core that can be readily ported on any Internet of Things (IoT) edge gateway that is connected to the DAQ.

Abstract: Systems and methods described herein can involve management of a system having a plurality of sensors, the plurality of sensors observing a plurality of process steps, which can involve selecting a subset of the plurality of sensors for observation; executing anomaly detection from data provided from the subset of the plurality of sensors; for a detection of an anomaly from a sensor from the subset of sensors, selecting ones of the plurality of process steps based on the detected anomaly; estimating a probability of anomaly occurrence for the selected ones of the plurality of process steps; and for the estimated probability of anomaly occurrence meeting a predetermined criteria, selecting ones of the plurality of sensors associated with the selected ones of the plurality of process steps for observation.

Abstract: Example implementations described herein involve systems and methods for training and managing machine learning models in an industrial setting. Specifically, by leveraging the similarity across certain production areas, example implementations can group together these areas to train models efficiently that use human pose data to predict human activities or specific task(s) the workers are engaged in. The example implementations do away with previous methods of independent model construction for each production area and takes advantage of the commonality amongst different environments.

Abstract: Example implementations described herein involve systems and methods to select machine learning models that will be executed in a cellular Mobile Edge Computing for cellular enabled applications. In contrast to related art implementations, the example implementations described herein considers different data fidelities of received data due to the cellular wireless channel and also performs service resource allocation accordingly.

Abstract: Example implementations described herein involve a system for training and managing machine learning models in an industrial setting. Specifically, by leveraging the similarity across certain production areas, it is possible to group such areas together to train models efficiently that use human pose data to predict human activities or specific task(s) that the workers are engaged in. Example implementations remove previous methods of independent model construction for each production area and takes advantage of the commonality amongst different environments.

Abstract: Example implementations described herein involve systems and methods to substantially simultaneously orchestrate machine learning models over multiple resource constrained control edge devices, so that the overall system is more agile to changes in events and environmental conditions where the models have been deployed. The example implementations described herein involve multiple processes that when executed, determine a list of edge devices to be updated along with the corresponding models based on correlation.

Abstract: Example implementations described herein can dynamically adapt to changing nature of sensor data traffic and through artificial intelligence (AI, strike a good tradeoff between reducing volume of sensed data, and retain enough data fidelity so that subsequent analytics applications perform well. The example implementations eliminate heuristic methods of setting sensing parameters (such as DAQ sampling rate, resolution etc.) and replaces them with an automated, AI driven edge solution core that can be readily ported on any Internet of Things (IoT) edge gateway that is connected to the DAQ.

Abstract: In example implementations described herein, the power of time series machine learning is used to extract the statistics of Programmable Logic Controller (PLC) data and external sensor data. The accuracy of time series machine learning is improved by manufacturing context-dependent segmentation of the time series into states which is factory may be in. The invention can capture subtle trends in these time series data and be able to classify them into several outcomes from ICS security attacks to normal anomalies and machine/sensor failures.

Abstract: Example implementations described herein are directed to systems and methods for non-invasive data extraction from digital displays. In an example implementation, a method includes receiving one or more video frames from a video capture device capturing an external display, where the external display is independent the video capture device; determining one or more locations within the external display comprising time varying data of the external display; and for each identified location of the time varying data: determining a data type; applying one or more rules based on the data type; and determining an accuracy of the time varying data within the one or more frames based on the rules.