Abstract: An intersecting road estimation device includes an object detection unit that detects an object existing around an own vehicle and a position of the object, an object extraction unit that extracts a stationary object and a position of the stationary object from an object detection result, a first estimation unit that estimates a road edge of a traveling road on which the own vehicle is traveling based on the position of the stationary object, a candidate extraction unit that extracts a stationary object existing outside the road edge of the traveling road estimated by the first estimation unit as a candidate for an outside stationary object representing a road edge of an intersecting road intersecting the traveling road, and a second estimation unit that estimates the road edge of the intersecting road based on a position of the outside stationary object extracted by the candidate extraction unit.

Abstract: Disclosed are techniques for compressing data in a data storage system comprising searching a cluster of nearest neighbors, wherein the cluster has been created using a locality sensitive hashing algorithm, to determine if a data block can be compressed. In alternate embodiments, nearest neighbor clusters can be formed using unsupervised learning. Additionally, nearest neighbors can also be formed in alternate embodiments using one or more of the following algorithms: a k-means clustering algorithm, a k-medoids clustering algorithm, a mean shift algorithm, a generalized method of moment (GMM) algorithm, or a density based spatial clustering of applications with noise (DBSCAN) algorithm.

Abstract: Provided is a method for implementing reinforcement learning by a neural network. The method may include performing, for each epoch of a first predetermined number of epochs, a second predetermined number of training iterations and a third predetermined number of testing iterations using a first neural network. The first neural network may include a first set of parameters, the training iterations may include a first set of hyperparameters, and the testing iterations may include a second set of hyperparameters. The testing iterations may be divided into segments, and each segment may include a fourth predetermined number of testing iterations. A first pattern may be determined based on at least one of the segments. At least one of the first set of hyperparameters or the second set of hyperparameters may be adjusted based on the pattern. A system and computer program product are also disclosed.

Abstract: Techniques and systems are provided for configuring neural networks to perform certain image manipulation operations. For instance, in response to obtaining an image for manipulation, an image manipulation system determines the fitness scores for a set of neural networks resulting from the processing of a noise map. Based on these fitness scores, the image manipulation system selects a subset of the set of neural networks for cross-breeding into a new generation of neural networks. The image manipulation system evaluates the performance of this new generation of neural networks and continues cross-breeding this neural networks until a fitness threshold is satisfied. From the final generation of neural networks, the image manipulation system selects a neural network that provides a desired output and uses the neural network to generate the manipulated image.

Abstract: The present disclosure relates to an image processing device including: a memory configured to store one or more instructions; and a processor configured to execute the one or more instructions stored in the memory to: extract one or more input patches based on an input image; extract one or more pieces of feature information respectively corresponding to the one or more input patches, based on a dictionary including mapping information indicating mappings between a plurality of patches and pieces of feature information respectively corresponding to the plurality of patches; and obtain a final image by performing a convolution operation between the extracted one or more pieces of feature information and a filter kernel.

Abstract: Certain embodiments of the disclosure relate to an apparatus and a method for translating a text included in an image by using an external electronic device in an electronic device. One method comprises displaying a picture comprising an object bearing text at a location within the picture on a display, extracting the text, generating another text from the extracted text, and automatically overlaying the another text on the object in another picture comprising the object at another location within the another picture on the display.

Abstract: A point cloud encoding method includes grouping a to-be-coded point cloud group into a plurality of subgroups, where the grouping a to-be-coded point cloud group into a plurality of subgroups includes pre-grouping a plurality of frames of point clouds in the to-be-coded point cloud group to obtain a pre-grouped subgroup, and determining, based on feature information of the pre-grouped subgroup, that the pre-grouped subgroup is one of the plurality of subgroups, where the feature information represents a size of an occupancy map of a point cloud in the pre-grouped subgroup, and encoding a point cloud included in the plurality of subgroups.

Abstract: An apparatus for estimating contact distribution to quickly estimate contact spot distribution from a contact surface image using a deep learning model based on a convolution neural network (CNN) and a method thereof are disclosed. A method for estimating contact distribution to estimate contact spot distribution between a first contact spot and a second contact spot includes inputting a contact surface image of at least one of the first contact surface and the second contact surface to a deep learning model based on a CNN and estimating contact spot distribution between the first contact surface and the second contact surface from an output of the deep learning model.

Abstract: A method of anomaly detection for network traffic communicated by devices via a computer network, the method including clustering a set of time series, each time series including a plurality of time windows of data corresponding to network communication characteristics for a device; training an autoencoder for each cluster based on time series in the cluster; generating a set of reconstruction errors for each autoencoder based on testing the autoencoder with data from time windows of at least a subset of the time series; generating a probabilistic model of reconstruction errors for each autoencoder; and generating an aggregation of the probabilistic models for, in use, detecting reconstruction errors for a time series of data corresponding to network communication characteristics for a device as anomalous.

Abstract: Systems and methods include processors for receiving training data for a user activity; receiving bias criteria; determining a set of model parameters for a machine learning model including: (1) applying the machine learning model to the training data; (2) generating model prediction errors; (3) generating a data selection vector to identify non-outlier target variables based on the model prediction errors; (4) utilizing the data selection vector to generate a non-outlier data set; (5) determining updated model parameters based on the non-outlier data set; and (6) repeating steps (1)-(5) until a censoring performance termination criterion is satisfied; training classifier model parameters for an outlier classifier machine learning model; applying the outlier classifier machine learning model to activity-related data to determine non-outlier activity-related data; and applying the machine learning model to the non-outlier activity-related data to predict future activity-related attributes for the user activity

Abstract: A method includes the steps of obtaining a frame from an image sensor, the frame comprising a number of pixel values, detecting a change in a first subset of the pixel values, detecting a change in the second subset of the pixel values near the first subset of the pixel values, and determining an occupancy state based on a relationship between the change in the first subset of the pixel values and the second subset of the pixel values. The occupancy state may be determined to be occupied when the change in the first subset of the pixel values is in a first direction and the change in the second subset of the pixel values is in a second direction opposite the first direction.

Abstract: A vehicle and a system and method of operating the vehicle based on a gesture made by a traffic director. The system includes a camera and at least one neural network. The camera obtains an image of a flag operator. The at least one neural network is to generates an encoded hand vector based on a configuration of a hand of the traffic director from the image, combines a skeleton of the traffic director generated from the image and the encoded hand vector to generate a representation vector, and predicts a gesture of the traffic director from the representation vector.

Abstract: A data processing method for a biochip comprises: acquiring a biochip image to be detected; performing binarization processing on the biochip image to obtain a binary image; performing a morphological dilation operation on the binary image in a row direction to obtain a first image, and performing a morphological dilation operation on the binary image in a column direction to obtain a second image; performing connected domain detection on the first image in the row direction, and performing connected domain detection on the second image in the column direction, to determine the number of rows and the number of columns of a sample point array and center position information of each sample point.

Abstract: Systems, methods, and other embodiments described herein relate to training a multi-task network using real and virtual data. In one embodiment, a method includes acquiring training data that includes real data and virtual data for training a multi-task network that performs at least depth prediction and semantic segmentation. The method includes generating a first output from the multi-task network using the real data and second output from the multi-task network using the virtual data. The method includes generating a mixed loss by analyzing the first output to produce a real loss and the second output to produce a virtual loss. The method includes updating the multi-task network using the mixed loss.

Abstract: A data processing method and device are provided. The method includes: extracting a plurality of data sets from unlabeled data; and for each data set, creating a plurality of sample sets by assigning labels to data samples in the data set, respectively training, for each sample set created from the data set, a classifier by using the sample set and labeled data, obtaining a sample set that corresponds to a trained classifier with the highest performance, and adding the obtained sample set to a candidate training set. Each sample set includes the first preset number of data samples with respective labels, the labels of the data samples in each sample set constitutes a label combination, and label combinations corresponding to different sample sets are different from each other. The method also includes adding a second preset number of sample sets in the candidate training set to the labeled data.

Abstract: A tracking device and method of determining a location of a tag connected to an object. The method comprises transmitting a first message from a server over a long range communication protocol to the tag so that the tag emits short and long range distress messages; receiving at a server the long range distress messages including location data using the long range communication protocol; using the location data to determine a geographical area in which the tag is located; transmitting a message from the server to one or more mobile devices in the geographic area to enable the mobile devices to receive the short range distress messages; receiving at the mobile devices the short range distress messages from the tag and using the received short range distress message to locate the tag; and transmitting a second message from the mobile devices to the server including the location of the tag.

Abstract: The present application provides a comparison method and a modeling method for a chip product, a device and a storage medium. According to the method, the chip product is modeled by using a neural network based on a slice sequence of the chip product in advance to obtain a three-dimensional stereoscopic model. When the chip products are compared, a comparison feature is acquired responsive to an operation of a user. For each chip product, a comparison result corresponding to the comparison feature is acquired from the three-dimensional stereoscopic model corresponding to each chip product. Then, the comparison result corresponding to each chip product is displayed.

Abstract: Described are techniques for training an image classifier using an augmentation loss function. The techniques including inputting corresponding pairs of a plurality of training images to an image classifier, where respective pairs of the corresponding pairs comprise at least two images having a same classification and different augmentations. The techniques further including training an artificial neural network of the image classifier to classify the plurality of training images using an augmentation loss function, wherein the augmentation loss function reduces differences in model outputs between the corresponding pairs of the plurality of training images.

Abstract: A dual autoencoder includes a first domain encoder for mapping a first domain image into a first latent space, a first domain decoder for reconstructing the first domain image, a second domain encoder for mapping a second domain image into a second latent space, a second domain decoder for reconstructing the second domain image; and a latent space association network for defining a cross-domain relationship between the first domain and the second domain. An image translation method using a dual encoder, may include: taking a first domain image as an input; determining an output condition; if the output condition is the same domain, reconstructing the first domain image by a first domain encoder and a first domain decoder; and if the output condition is the cross domain, reconstructing a second domain image by the first domain encoder, a latent space association network, and a second domain decoder.

Abstract: A machine learning model is described that is trained without labels to predict a motion field between a pair of images. The trained model can be applied to a distinguished pair of images to predict a motion field between the distinguished pair of images.