Abstract: Embodiments of the disclosure provide a machine learning model for generating a predicted executable command for an image. The learning model includes an interface configured to obtain an utterance indicating a request associated with the image, an utterance sub-model, a visual sub-model, an attention network, and a selection gate. The machine learning model generates a segment of the predicted executable command from weighted probabilities of each candidate token in a predetermined vocabulary determined based on the visual features, the concept features, current command features, and the utterance features extracted from the utterance or the image.

Abstract: Range adaptive, dynamic metadata generation for high-dynamic images includes generating, using computer hardware, histogram-based data for video including one or more frames. The histogram-based data is generated for each of a plurality of dynamic ranges. For each dynamic range, a predetermined amount of dynamic metadata for the video is generated from the histogram-based data for the dynamic range. The video and the dynamic metadata is output.

Abstract: To enable end-to-end performance monitoring, different solutions to obtain end-to-end values for different key performance indicators are disclosed. For example, when key performance indicator data from a plurality of different domains is received, the key performance indicator data received may be adapted to be a time-series data stream. Then, values of key performance indicators may be extracted, per a domain, from the time-series data stream. An end-to-end value for a key performance indicator may be determined, based on the extracted values of the key performance indicator in the different domains. The end-to-end value may be used for end-to-end performance monitoring of a network.

Abstract: A method for making light emitting device LED arrays includes the steps of providing a plurality of LEDs having a desired configuration (e.g., VLED, FCLED, PLED); attaching the LEDs to a carrier substrate and to a temporary substrate; forming one or more metal layers and one or more insulator layers configured to electrically connect the LEDs to form a desired circuitry; and separating the LEDs along with the layered metal layers and insulator layers that form the desired circuitry from the carrier substrate and the temporary substrate.

Abstract: Electric heaters for heating a hydrocarbon process stream. The electric heater has a cavity with parallel walls which may provide a rectangular cross section to the cavity. The parallel side walls may be parallel to a longitudinal axis of the electrical heating elements which extend into the cavity to transfer heat to fluid passing there through. Various configurations and orientations of the electrical heating elements are provided.

Abstract: Processes and apparatuses for heating a hydrocarbon process stream, with an electric heater to provide a portion of the heat requirement necessary for a chemical reaction to occur to one of the components of the hydrocarbon process stream. The electric heater may be used between two reaction zones which are arranged on top of each other. Additionally, reaction zones may be in reactors which are arranged in a polygon.

Abstract: In various examples, a three-dimensional (3D) intersection structure may be predicted using a deep neural network (DNN) based on processing two-dimensional (2D) input data. To train the DNN to accurately predict 3D intersection structures from 2D inputs, the DNN may be trained using a first loss function that compares 3D outputs of the DNN—after conversion to 2D space—to 2D ground truth data and a second loss function that analyzes the 3D predictions of the DNN in view of one or more geometric constraints—e.g., geometric knowledge of intersections may be used to penalize predictions of the DNN that do not align with known intersection and/or road structure geometries. As such, live perception of an autonomous or semi-autonomous vehicle may be used by the DNN to detect 3D locations of intersection structures from 2D inputs.

Abstract: A method for measuring an amount of a gaseous species present in a gas, the gaseous species absorbing light in an absorption spectral band, comprises placing the gas between a light source and a measuring photodetector. The light source is configured to emit a light wave that propagates through the gas to the measuring photodetector. The light source is activated so as to illuminate the gas, so that the light source emits a light pulse. The method also includes measuring, with the measuring photodetector, a measurement intensity of a light wave transmitted by the gas during the illumination, in a measurement spectral band. The measurement spectral band comprises the absorption spectral band. The light source is activated using a pulsed activation signal, each pulse having a specific form, notably to reduce aging of the source.

Abstract: A method and a system for performing distance metric learning using proxies are provided. The method for performing distance metric learning assigns a proxy as an anchor to represent a class and associates the proxy with all data points in a training batch. The method allows data points to interact with each other via proxies during training. Additionally, the fine-grained data-to-data relation is actively considered, which is combined with a learnable margin parameter leading to intra-class compactness and inter-class separability.

Abstract: Systems and methods for video segmentation and summarization are described. Embodiments of the present disclosure receive a video and a transcript of the video; generate visual features representing frames of the video using an image encoder; generate language features representing the transcript using a text encoder, wherein the image encoder and the text encoder are trained based on a correlation between training visual features and training language features; and segment the video into a plurality of video segments based on the visual features and the language features.

Abstract: An electromechanical power switch device and methods thereof. At least some of the illustrative embodiments are devices including a semiconductor substrate, at least one integrated circuit device on a front surface of the semiconductor substrate, an insulating layer on the at least one integrated circuit device, and an electromechanical power switch on the insulating layer. By way of example, the electromechanical power switch may include a source and a drain, a body region disposed between the source and the drain, and a gate including a switching metal layer. In some embodiments, the body region includes a first body portion and a second body portion spaced a distance from the first body portion and defining a body discontinuity therebetween. Additionally, in various examples, the switching metal layer may be disposed over the body discontinuity.

Abstract: An asset monitoring system receives a first multi-dimensional representation of an asset associated with an industrial process. The asset monitoring system generates a second multi-dimensional representation of the asset from the first multi-dimensional representation. The second multi-dimensional representation may be a two-dimensional vector graphic representation of the asset and may be configurable via a graphical user interface. In generating the second multi-dimensional representation, the asset monitoring system maps at least one operational parameter associated with the asset to the second multi-dimensional representation. The asset may include one or more devices of the asset monitoring system. The asset monitoring system may display, via a dashboard at the graphical user interface, the second multi-dimensional representation and status information associated with the asset.

Abstract: The invention relates to a method of calculating gating scores in the pulse-Doppler radar using a combination of position, power, size of detection and ambiguous Doppler measurement data to effectively overcome the phenomenon of wrongly assigning detections to the target trajectory and creating false targets in noise regions. The method intelligently merges position, power, detection size and ambiguous Doppler measurement data information of the detections in the algorithm of calculating gating score to solve the problem of selection and disputation detections. The method is carried out through two steps, step 1: determine statistical quantities from the standard data set; step 2: gating selection algorithm.

Abstract: Techniques for using more-specific routing to perform scalable Layer-2 (L2) stretching of subnets across hybrid-cloud environments. Routing tables in a public cloud may allow for routes that are more specific than the default local route, and the more-specific routes may be used to send all traffic to a dedicated, cloud router. The more-specific routes are set up for a VPC where a subnet resides such that the more specific-routes cover at least a portion of subnet range. The next hop for the more-specific routes point to the cloud router which is capable of doing host routing and segmentation extension. Thus, traffic originating from endpoints in a VPC is routed to the cloud router, and the cloud router determines whether the traffic is to be re-routed back to a destination endpoint in the VPC (or another cloud location), or sent to a destination endpoint residing in the on-premises site.

Abstract: A computer system comprising a processor and a memory storing instructions that, when executed by the processor, cause the computer system to perform a set of operations. The set of operations comprises collecting domain attribute data comprising one or more domain attribute features for a domain, collecting sampled domain profile data comprising one or more domain profile features for the domain and generating, using the domain attribute data and the sampled domain profile data, a domain reputation assignment utilizing a neural network.

Abstract: Aspects of the present disclosure provide for a patient warming system having a heating device configured to transfer heat to a patient based on a body heat score that indicates whether the patient is likely to sweat. The heating device includes a heater circuit configured to produce heat, the heater circuit thermally coupled to the patient. The heating device also includes a controller communicatively coupled to the heater circuit. The controller has one or more processors configured to receive one or more physiological indicators of the patient, determine the body heat score from one or more physiological indicators, determine whether the body heat score meets a score threshold, direct the heater circuit to produce a first amount of heat based on whether the body heat score meets a score threshold.

Abstract: Described herein are methods, compositions and articles of manufacture involving neutral conjugated polymers including methods for synthesis of neutral conjugated water-soluble polymers with linkers along the polymer main chain structure and terminal end capping units. Such polymers may serve in the fabrication of novel optoelectronic devices and in the development of highly efficient biosensors. The invention further relates to the application of these polymers in assay methods.

Abstract: A device may access a database of devices, wherein the database comprises a plurality of templates, and wherein each of the plurality of templates comprises asset-related information and a plurality of predefined algorithms. The device may provide a graphical user interface for a user to input information related to a targeted device. The device may automatically select a targeted operation template among the plurality of templates for the asset in response to either the input information or a change of condition of a neighboring device connected to the targeted device. The device may automatically modify one of the plurality of predefined algorithms that is associated with the targeted operation template based on a peripheral device associated with the asset for an optimization purpose. The device may calculate asset performance indicators as a measure to determine how far the optimization purpose has been achieved.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for generating aspect-based summaries utilizing deep learning. In particular, in one or more embodiments, the disclosed systems access a transcript comprising sentences. The disclosed systems generate, utilizing a sentence classification machine learning model, aspect labels for the sentences of the transcript. The disclosed systems organize the sentences based on the aspect labels. The disclosed systems generate, utilizing a summary machine learning model, a summary of the transcript for each aspect of the plurality of aspects from the organized sentences.

Abstract: There are provided a method, an apparatus, a device, and a storage medium for model training. In a method, a target model is fine-tuned using a set of training data, each training data including a sample question and corresponding annotation information, the annotation information including policy information for solving the sample question and answer information of the sample question. At least one sample question in the set of training data is provided to the fine-tuned target model to determine a candidate answer to the at least one sample question. The fine-tuned target model is trained based at least on a comparison between the candidate answer and the answer information of the at least one sample question.