Abstract: A vehicle control device is disclosed. The vehicle control device includes: an interface unit that is connected to a camera for capturing a vehicle passenger and to a plurality of devices installed inside the vehicle; and a processor that receives an image captured by the camera through the interface unit and exchanges information with the plurality of devices through the interface unit, wherein the processor detects the position of the passenger from the image acquired by the camera, configures a space adjacent to the passenger as a virtual control space, matches a device corresponding to the position of the passenger, among the plurality of devices, to the virtual control device, and, upon detecting a preset gesture through the virtual control space, controls the device based on the preset gesture.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: In one embodiment, a method includes receiving a user input from a user from a client system associated with the user, wherein the client system comprises one or more cameras, determining one or more points of interest in a field of view of the one or more cameras based on one or more machine-learning models and sensory data captured by the one or more cameras, generating a plurality of media files based on the one or more points of interest, wherein each media file is a recording of at least one of the one or more points of interest, generating one or more highlight files based on the plurality of media files, wherein each highlight file comprises a media file that satisfies a predefined quality standard, and sending instructions for presenting the one or more highlight files to the client system.

Abstract: Methods and systems for analyzing an integrity of a roof covering are presented. One or more roof sections of a structure may be identified with one or more processors. An airflow may then be directed over one or more shingles or other discontinuous roof covering materials of the roof section by an airflow unit. A respective lift for each of the shingles may then be observed by an imaging unit. An adhesion score may then be determined for each of the one or more shingles using the one or more processors. This adhesion score may be based on the respective lift. An overall roof integrity rating may then be generated by the one or more processors, based on the one or more adhesion scores. The overall roof integrity rating may then be provided by the one or more processors to a user of a computing device.

Abstract: A goods receipt management system includes processors coupled to a non-transitory storage device and operable to execute routines that include a face recognition engine to use facial features of a user to recognize the user as an authorised person, a dialogue manager engine to obtain from the authorised person a stated number of goods items being delivered, a quantification engine to receive an image of the goods items and calculate the number of goods items appearing in the image, a comparison engine to compare the calculated number of goods items with the stated number of goods items, and in the event of a substantial match, to record the calculated number of goods items, and a performance improvement engine to use the image of the goods items to re-tune configuration variables, in the event of a mis-match between calculated number of goods items and stated number of goods items.

Abstract: A processor-implemented recognition method includes: receiving query input data; determining a domain to which the query input data belongs using a neural network-based classifier; and in response to the query input data belonging to a first domain, generating second query data of a second domain based on the query input data.

Abstract: A technical document scanner disclosed herein determines and categorizes various common issues among a large number of documents. An implementation of the technical document scanner is implemented using various computer process instructions including scanning a technical document to extract content, applying named entity recognition on the extracted content from the technical document to extract named entities, applying relation extraction on the named entities to extract relations between the named entities, and analyzing the relations between the entities to compose lists of high relevance entities for issue checking.

Abstract: In an approach for detecting key messages for a video, a processor builds a role model based on data from one or more data sources, with an identification feature of each role in a video. A processor samples a plurality of frames of the video. A processor identifies a key object presented in the plurality of frames. The key object is a role in the video. A processor recognizes a movement scenario associated with the role. A processor dynamically updates the role model based on the movement scenario. A processor identifies a role name based on the movement scenario. A processor generates a description script associated with the movement scenario for the role. A processor outputs the description script.

Abstract: The present specification discloses systems and methods for identifying and reporting contents of a tanker, container or vehicle. Programmatic tools are provided to assist an operator in analyzing contents of a tanker, container or vehicle. Manifest data is automatically imported into the system for each shipment, thereby helping security personnel to quickly determine container contents. In case of a mismatch between container contents shown by manifest data and the contents as ascertained from the scanning system, the container or vehicle may be withheld for further inspection.

Abstract: Systems and methods are provided for performing a video-grounded dialogue task by a neural network model using bi-directional spatial-temporal reasoning. According to some embodiments, the systems and methods implement a dual network architecture or framework. This framework includes one network or reasoning module that learns dependencies between text and video in the direction of spatial?temporal, and another network or reasoning module that learns in the direction of temporal?spatial. The output of the multimodal reasoning modules may be combined to learn dependencies between language features in dialogues. The result joint representation is used as a contextual feature to the decoding components which allow the model to semantically generate meaningful responses to the users. In some embodiments, pointer networks are extended to the video-grounded dialogue task to allow the model to point to specific tokens from multiple source sequences to generate responses.

Abstract: A speech recognition method and apparatus, including implementation and/or training, are disclosed. The speech recognition method includes obtaining a speech signal, and performing a recognition of the speech signal, including generating a dialect parameter, for the speech signal, from input dialect data using a parameter generation model, applying the dialect parameter to a trained speech recognition model to generate a dialect speech recognition model, and generating a speech recognition result from the speech signal by implementing, with respect to the speech signal, the dialect speech recognition model. The speech recognition method and apparatus may perform speech recognition and/or training of the speech recognition model and the parameter generation model.

Abstract: Deep recurrent neural networks applied to speech recognition. The deep recurrent neural networks (RNNs) are preferably implemented by stacked long short-term memory bidirectional RNNs. The RNNs are trained using end-to-end training with suitable regularisation.

Abstract: Certain embodiments involve techniques for efficiently estimating denoising kernels for generating denoised images. For instance, a neural network receives a noisy reference image to denoise. The neural network uses a kernel dictionary of base kernels and generates a coefficient vector for each pixel in the reference image such that the coefficient vector includes a coefficient value for each base kernel in the kernel dictionary, where the base kernels are combined to generate a denoising kernel and each coefficient value indicates a contribution of a given base kernel to a denoising kernel. The neural network calculates the denoising kernel for a given pixel by applying the coefficient vector for that pixel to the kernel dictionary. The neural network applies each denoising kernel to the respective pixel to generate a denoised output image.

Abstract: Embodiments of the present application provide a method, apparatus and device for interaction of intelligent voice devices, and a storage medium. The method includes: receiving wake-up messages sent by respective awakened intelligent voice devices; determining a forwarding device according to the wake-up messages; sending a forwarding instruction to the forwarding device to enable the forwarding device to receive a user voice request according to the forwarding instruction, where the forwarding instruction includes: type skill information of all intelligent voice devices; and sending a non-response message to other awakened intelligent voice device other than the forwarding device, which enables the most appropriate response device to execute the cloud result requested by the forwarding device, and the plurality of awakened intelligent voice devices do not respond at the same time so as to avoid confusion, thus making it easier to meet user needs.

Abstract: A method for generating training data can include: determining a set of images; determining a set of masks based on the images; determining a first mesh based on the set of masks; optionally determining a refined mesh by recomputing the first mesh; optionally determining one or more faces of the refined mesh; optionally adding one or more keypoints to the refined mesh; optionally determining a material property set for the object; optionally generating a full object mesh; determining one or more scenes; optionally determining training data based on the one or more scenes; optionally training one or more object detectors using the training data; and detecting one or more objects using the trained object detector.

Abstract: A trained model is described in the form of a Bayesian neural network (BNN) which provides a quantification of its inference uncertainty during use and which is trained using marginal likelihood maximization. A Probably Approximately Correct (PAC) bound may be used in the training to incorporate prior knowledge and to improve training stability even when the network architecture is deep.

Abstract: Techniques for determining predictions on a top-down representation of an environment based on object movement are discussed herein. Sensors of a first vehicle (such as an autonomous vehicle) may capture sensor data of an environment, which may include object(s) separate from the first vehicle (e.g., a vehicle, a pedestrian, a bicycle). A multi-channel image representing a top-down view of the object(s) and the environment may be generated based in part on the sensor data. Environmental data (object extents, velocities, lane positions, crosswalks, etc.) may also be encoded in the image. Multiple images may be generated representing the environment over time and input into a prediction system configured to output a trajectory template (e.g., general intent for future movement) and a predicted trajectory (e.g., more accurate predicted movement) associated with each object. The prediction system may include a machine learned model configured to output the trajectory template(s) and the predicted trajector(ies).

Abstract: In some embodiments, a method inputs a set of images into a network and trains the network based on a classification of the set of images to one or more characters in a set of characters. The method obtains a set of encodings for the one or more characters based on a layer of the network that restricts the output of the layer to a number of values. Then, the method stores the set of encodings for the one or more characters, wherein an encoding in the set of encodings is retrievable when a corresponding character is determined.

Abstract: In order to improve a user's task of documenting items that are identified by the user while at a construction site, a computing system may be configured to (i) analyze media content captured at a construction site (e.g., an audio or audiovisual recording) that is descriptive of at least one given on-site item identified by a user while at the construction site, (ii) extract, from the captured media content, a set of information for the given on-site item, (iii) based on the extracted set of information for the given on-site item, generate a data record for the given on-site item, and (iv) cause the generated data record for the given on-site item to be presented to the user for review.

Abstract: Disclosed are a method and an apparatus for recommending clothing coordination information in the 5G communication environment by executing an artificial intelligence (AI) algorithm and/or a machine learning algorithm, which have been mounted therein.