Abstract: An effective diagnosis region detection unit uses an effective diagnosis region detection model and detects an effective diagnosis region in which a non-target region of interest other than a region of interest is removed from an observation target, from a first medical image. The effective diagnosis region detection model is obtained by learning using first learning data including the first medical image and effective diagnosis region information regarding the effective diagnosis region. The region-of-interest detection unit detects the region of interest from the effective diagnosis region.

Abstract: Implementations are directed to generating simulated training examples for training of a machine learning model, training the machine learning model based at least in part on the simulated training examples, and/or using the trained machine learning model in control of at least one real-world physical robot. Implementations are additionally or alternatively directed to performing one or more iterations of quantifying a “reality gap” for a robotic simulator and adapting parameter(s) for the robotic simulator based on the determined reality gap. The robotic simulator with the adapted parameter(s) can further be utilized to generate simulated training examples when the reality gap of one or more iterations satisfies one or more criteria.

Abstract: Methods, apparatus, systems and articles of manufacture are disclosed to map workloads. An example apparatus includes a constraint definer to define performance characteristic targets of the neural network, an action determiner to apply a first resource configuration to candidate resources corresponding to the neural network, a reward determiner to calculate a results metric based on (a) resource performance metrics and (b) the performance characteristic targets, and a layer map generator to generate a resource mapping file, the mapping file including respective resource assignments for respective corresponding layers of the neural network, the resource assignments selected based on the results metric.

Abstract: A computer-implemented method for analyzing a survivability of a structure is presented. A distance extending radially from the structure defining an area surrounding the structure may be determined. The area surrounding the structure may be scanned to identify surrounding objects. The area surrounding the structure may then be segmented to create a plurality of area segments. A density of surrounding objects in each of the plurality of area segments may be determined. A risk score for each of the plurality of area segments may then be determined based on the density of surrounding objects in each of the plurality of area segments. An overall wind exposure risk value may be provided for further use.

Abstract: A method for predicting an air quality index (AQI) based on a fusion model proposes a differential fusion seasonal prediction model (DF-SPM) based on a random forest (RF) model and a convolutional neural network (CNN)-long short-term memory (LSTM)-attention (CLA) model. This method uses the optimal threshold interval (OTI) search algorithm to search and learn the OTI of four seasons during the prediction process, and obtains the final prediction results according to the predicted values of RF model and CLA model. The fusion model combines the prediction advantages of two independent models, and fully considers the seasonal and periodic characteristics of AQI, so as to accurately search OTI in different time periods with the seasonal scale, so as to achieve higher prediction accuracy. The OTI strategy of fusion model is superior to the single threshold strategy, which can extract the historical fluctuation characteristics of AQI and achieve higher prediction accuracy.

Abstract: An action recognition apparatus 1 including: a generation unit 2 that generates an addition image by adding a setting image to a target object image; an action recognition and estimation unit 3 that performs action recognition regarding the target object, outputs action recognition information indicating the result of action recognition, estimates the setting image, and outputs estimation information indicating the result of estimation; an action recognition loss calculation unit 4 that calculates an action recognition loss by using reference action recognition information generated in advance, and the action recognition information; an estimation loss calculation unit 5 that calculates an estimation loss by using reference estimation information generated in advance, and the estimation information; and a learning parameter updating unit 6 that updates learning parameters by using the action recognition loss and the estimation loss.

Abstract: A method includes accessing a web-based property over a network; storing a plurality of images or videos from the web-based property and associations between the plurality of images or videos and a target audience identifier responsive to the web-based property having a stored association with the target audience identifier; retrieving the plurality of images or videos from the database responsive to each of the plurality of images or videos having stored associations with the target audience identifier; executing a neural network to generate a performance score for each of the plurality of images or videos; calculating a target audience benchmark; executing the neural network to generate a first performance score for a first image or video and a second performance score for a second image or video; comparing the first performance score and the second performance score to the benchmark; and generating a record identifying the first image or video.

Abstract: A method for generating labeling data is disclosed that describes an image content of images depicting at least one scene, wherein in a processing device image data are received from an imaging and a segmentation unit that detects at least one object in the image data. A graphical processing unit generates a respective graphical object marker that marks the at least one detected object and a display control unit displays an overlay of the at least one scene and the at least one object marker. An input reception unit receives a respective user input for each object marker, wherein the respective user input provides the image content of the image region marked by the object marker.

Abstract: Generally discussed herein are devices, systems, and methods for on-device detection of a wake word. A device can include a memory including model parameters that define a custom wake word detection model, the wake word detection model including a recurrent neural network transducer (RNNT) and a lookup table (LUT), the LUT indicating a hidden vector to be provided in response to a phoneme of a user-specified wake word, a microphone to capture audio, and processing circuitry to receive the audio from the microphone, determine, using the wake word detection model, whether the audio includes an utterance of the user-specified wake word, and wake up a personal assistant after determining the audio includes the utterance of the user-specified wake word.

Abstract: Methods, apparatus and systems related to automated evaluation of acting performance are disclosed. In one example aspect, a system for an automated evaluation of performance activities includes a user interface configured to allow a user to create a customized automated evaluation service. The user can select a subset of machine learning engines from one or more available machine learning engines to create such service. The user interface is further configured to receive a set of input data of a performance performed by a performer. The system includes the one or more available machine learning engines each configured to generate a sub-score in a specific area using the set of input data received from the user interface. The system also includes an aggregator configured to produce an output indicating a quality of the performance.

Abstract: The disclosure relates to a system and method for reconstructing a PET image. The method may include: obtaining PET data relating to an object collected by a plurality of detector units; determining functional status of the plurality of detector units; generating reconstruction data based on the functional status of the respective detector units and the PET data; and reconstructing a PET image based on the reconstruction data.

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: Provided is an information processing system including: a comparison information acquisition unit that acquires comparison information regarding iris comparison generated based on an iris image including an iris of a recognition subject; and a display image generation unit that generates a display image including an image indicating a content of the comparison information in association with positions in the iris.

Abstract: Methods and systems for analyzing an integrity of a roof covering are presented. An airflow may be directed over one or more shingles or other discontinuous roof covering materials by an airflow unit. A respective lift for each of the discontinuous roof covering materials may then be observed by an imaging unit, and an overall roof integrity rating may then be generated by the one or more processors based on the respective lift of the discontinuous roof covering materials.

Abstract: Provided is an information processing system including: a comparison information acquisition unit that acquires comparison information regarding iris comparison generated based on an iris image including an iris of a recognition subject; and a display image generation unit that generates a display image including an image indicating a content of the comparison information in association with positions in the iris.

Abstract: Provided is an information processing system including: a comparison information acquisition unit that acquires comparison information regarding iris comparison generated based on an iris image including an iris of a recognition subject; and a display image generation unit that generates a display image including an image indicating a content of the comparison information in association with positions in the iris.

Abstract: Provided is an information processing system including: a comparison information acquisition unit that acquires comparison information regarding iris comparison generated based on an iris image including an iris of a recognition subject; and a display image generation unit that generates a display image including an image indicating a content of the comparison information in association with positions in the iris.

Abstract: Example methods and systems for tomographic data analysis are provided. One example method may comprise: obtaining first three-dimensional (3D) feature volume data and processing the first 3D feature volume data using an AI engine that includes multiple first processing layers, an interposing forward-projection module and multiple second processing layers. Example processing using the AI engine may involve: generating second 3D feature volume data by processing the first 3D feature volume data using the multiple first processing layers, transforming the second 3D volume data into 2D feature data using the forward-projection module and generating analysis output data by processing the 2D feature data using the multiple second processing layers.

Abstract: Methods and apparatus for end-to-end document image quality assessment without having ground truth for characters. An image quality assessment device integrates end-to-end machine learning without using ground truth to automatically predict an overall quality scores for images of documents. The image quality assessment device augments the size of individual image patches extracted from the images based on the contents and size of the images to maintain sufficient size of training data without compromising the reliability of the predicted scores for the images.

Abstract: A method includes storing a database comprising a plurality of pointers to web pages and identifiers of entities associated with the plurality of pointers; receiving a first request comprising a first identifier; identifying subset of the plurality of pointers from the database responsive to each pointer of the subset having a stored association with a first identification that matches the first identifier; responsive to identifying the subset of the plurality of pointers, establishing, via one or more pointers, a connection with a server hosting a set of web pages associated with the subset of the plurality of pointers; retrieving one or more images or videos from each of the set of web pages over the established connection; calculating a performance score for each of the one or more images or videos; and generating a record identifying the performance score for each of the one or more images or videos.