Abstract: Technology is provided for generating conversation features for recorded conversations. The technology includes, receiving videos depicting a multiple-user interaction, segmenting the videos into multiple utterances based on identifying utterances from individual users, receiving label data for the utterance segments specifying conversation features, and storing the label data in association with the utterance segments.

Abstract: A method for testing a portable electronic device. The method comprises the following steps: acquiring at least one set of parameters; storing the at least one set of parameters in a first area of a storage unit of the portable electronic device; storing the at least one set of parameters in a second area of the storage unit; generating at least a first image by means of a first processing operation; in a first displaying step, displaying the at least one first image for a predetermined time period; in a first erasing step, erasing first intermediate results corresponding to calculation steps carried out during the first processing operation; generating at least one second image by means of a second processing operation; and in a second displaying step, displaying the at least one second image for a predetermined time period.

Abstract: An apparatus includes a tensor compute cluster having a plurality of tensor compute units to process a plurality of sub-feature maps in a machine learning application and a tensor memory cluster having a plurality of tensor feature map memory units to store the plurality of sub-feature maps. The apparatus also includes circuitry to partition an input feature map into the plurality of sub-feature maps such that sparsity in each of the plurality of sub-feature maps satisfies a predetermined threshold, and assign each of the plurality of sub-feature maps to one of the plurality of tensor compute units and one of the plurality of tensor feature map memory units for processing in parallel.

Abstract: Disclosed are systems and methods for generating data sets for training deep learning networks for key point annotations and measurements extraction from photos taken using a mobile device camera. The method includes the steps of receiving a 3D scan model of a 3D object or subject captured from a 3D scanner and a 2D photograph of the same 3D object or subject at a virtual workspace. The 3D scan model is rigged with one or more key points. A superimposed image of a pose-adjusted and aligned 3D scan model superimposed over the 2D photograph is captured by a virtual camera in the virtual workspace. Training data for a key point annotation DLN is generated by repeating the steps for a plurality of objects belonging to a plurality of object categories. The key point annotation DLN learns from the training data to produce key point annotations of objects from 2D photographs captured using any mobile device camera.

Abstract: A baseline event detection system to detect events by performing operations that include: generating a baseline data set; accessing a data stream; performing a comparison of the baseline data set and the data stream; and detecting an event based on the comparison.

Abstract: A control method for traffic lights includes: acquiring an image at a preset time point before a red light is turned off, wherein the image includes at least one vehicle waiting for the red light; determining a time length according to the image, wherein the time length is generated by an artificial intelligence algorithm; and controlling a green light according to the time length.

Abstract: Disclosed are methods and systems for using artificial intelligence (AI) for image recognition by using predefined coordinates to extract a portion of a received image, the extracted portion comprising a word to be identified having at least a first letter and a second letter; executing an image recognition protocol to identify the first letter; when the server is unable to identify the second letter, the server executes an AI model having a nodal data structure to identify the second letter based upon the identified first letter, the nodal data structure comprising a set of nodes where each node represents a letter, each node connected to at least one other node, wherein connection of a first node to a second node corresponds to a probability that a letter corresponding to the second node is used in a word subsequent to a letter corresponding to the first node.

Abstract: A heterogeneous sensor fusion apparatus includes a point processor configured to detect a first object by processing a detection point input from a first sensor, an image processor configured to detect a second object by processing an image input from a second sensor, a point-matching unit configured to calculate a matching rate by matching the detection point with the image, and to determine whether the first object and the second object are identical based on the calculated matching rate, an association unit configured to generate track information by fusing information from the first sensor and the second sensor when the first object and the second object are identical, and an output unit configured to output the generated track information.

Abstract: A method of training and using a machine learning model that controls for consideration of undesired factors which might otherwise be considered by the trained model during its subsequent analyses of new data. For example, the model may be a neural network trained on a set of training images to evaluate an insurance applicant based upon an image or audio data of the insurance applicant as part of an underwriting process to determine an appropriate life or health insurance premium. The model is trained to probabilistically correlate an aspect of the applicant's appearance with a personal and/or health-related characteristic. Any undesired factors, such as age, sex, ethnicity, and/or race, are identified for exclusion. The trained model receives the image (e.g., a “selfie”) of the insurance applicant, analyzes the image without considering the identified undesired factors, and suggests the appropriate insurance premium based only on the remaining desired factors.

Abstract: One variation of a method for monitoring manufacture of assembly units includes: receiving selection of a target location hypothesized by a user to contain an origin of a defect in assembly units of an assembly type; accessing a feature map linking non-visual manufacturing features to physical locations within the assembly type; for each assembly unit, accessing an inspection image of the assembly unit recorded by an optical inspection station during production of the assembly unit, projecting the target location onto the inspection image, detecting visual features proximal the target location within the inspection image, and aggregating non-visual manufacturing features associated with locations proximal the target location and representing manufacturing inputs into the assembly unit based on the feature map; and calculating correlations between visual and non-visual manufacturing features associated with locations proximal the target location and the defect for the set of assembly units.

Abstract: Methods, apparatus, systems and articles of manufacture are disclosed. An example apparatus includes an audio detector to determine a first audience count based on signatures of audio data captured in the media environment, a thermal image detector to determine a heat blob count based on a frame of thermal image data captured in the media environment, and an audience image detector to identify at least one audience member based on a comparison of a frame of audience image data with a library of reference audience images, the audience image detector to perform the comparison in response to the first audience count not matching the heat blob count.

Abstract: Authentication of a user is based on an electronic contact lens that contains a retinal scanner. The retinal scanner captures a retinal scan, which is used to authenticate the user (i.e., what a person is). The contact lens itself may also be used in the authentication process (i.e., what a person has). Authentication based on these two factors are used to control access to assets.

Abstract: An electronic device, to detect an object, may continuously obtain a driving image and convert the driving image into a top-view image, generate a change value including a horizontal change value based on a steering angle and a vertical change value based on a vehicle speed, compensate the top-view image based on a plurality of lanes to generate first data and compensate the first data based on the change value to generate second data, generate third data based on performing a subtraction operation on a first frame and a second frame of the second data, and a group the third data and extract a grouped object.

Abstract: An information processing apparatus includes: a first preprocessing arithmetic device configured to execute preprocessing for analog data from a first sensor; and a first post-processing arithmetic device connected to the first preprocessing arithmetic device and configured to execute post-processing for first preprocessed data, wherein the first preprocessing arithmetic device includes a first processor configured to: receive the analog data from the first sensor and convert the analog data into digital data; output feature data on the basis of a result of execution of feature extraction processing for the digital data; and output the feature data, and the first post-processing arithmetic device includes a second processor configured to: input the feature data; store the feature data in a first memory; and store, in the first memory, recognition result data based on a result of execution of recognition processing for the feature data.

Abstract: A system associated with predicting segmentation quality of segmented objects implemented in the analysis of copious image data is disclosed. The system receives a collection of image data related to a particular type of data. The image data is segmented into segmented data portions based on an object associated with the collection of image data. Regions of interest associated with the segmented data portions are determined. The quality of segmentation of the segmented data portions is determined for respective classification of the regions of interest. A classification label is assigned to the regions of interest. Regions of interest are partitioned into sub-regions. Features associated with the sub-regions of the segmented data portions are determined. A training dataset is generated based on the determined features associated with the sub-regions in order to train a classification model based on a predetermined threshold value.

Abstract: Systems, devices, and methods related to video analysis using an Artificial Neural Network (ANN) are described. For example, a data storage device can be configured to perform the computation of an ANN to recognize or classify features captured in the video images. The recognition or classification results of a prior video frame can be used to accelerate the analysis of the next video frame. The ANN can be organized in layers, where the intermediate result of a current layer can be further analyzed by a next layer for improved accuracy and confidence level. Before or while processing using the next layer, the intermediate result can be compared to the results obtained for the prior frame. If, in view of the results of the prior frame, the confidence level of the intermediate result is boosted to above a threshold, the subsequent layer(s) can be skipped or terminated early.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating object detection predictions from a neural network. In some implementations, an input characterizing a first region of an environment is obtained. The input includes a projected laser image generated from a three-dimensional laser sensor reading of the first region, a camera image patch generated from a camera image of the first region, and a feature vector of features characterizing the first region. The input is processed using a high precision object detection neural network to generate a respective object score for each object category in a first set of one or more object categories. Each object score represents a respective likelihood that an object belonging to the object category is located in the first region of the environment.

Abstract: Methods for automatically tagging one or more images and/or video clips using a audio stream are disclosed. The audio stream may be processed using an automatic speech recognition algorithm, to extract possible keywords. The image(s) and/or video clip(s) may then be tagged with the possible keywords. In some embodiments, the image(s) and/or video clip(s) may be tagged automatically. In other embodiments, a user may be presented with a list of possible keywords extracted from the audio stream, from which the user may then select to manually tag the image(s) and/or video clip(s).

Abstract: Versions of the disclosure relate to methods of imaging and detecting germinated seeds on a soilless growth medium.

Abstract: Systems and techniques for facilitating hand activity sensing are presented. In one example, a system obtains, from a wrist-worn computational device, hand activity data associated with a sustained series of hand motor actions in performance of a human task. The system also employs a machine learning technique to determine classification data indicative of a classification for the human task.