The present disclosure relates to systems, methods, and non-transitory computer readable media for training a classification neural network to classify digital images in few-shot tasks based on self-supervision and manifold mixup. For example, the disclosed systems can train a feature extractor as part of a base neural network utilizing self-supervision and manifold mixup. Indeed, the disclosed systems can apply manifold mixup regularization over a feature manifold learned via self-supervised training such as rotation training or exemplar training. Based on training the feature extractor, the disclosed systems can also train a classifier to classify digital images into novel classes not present within the base classes used to train the feature extractor.

A computer answers a question using a data table. The computer receives a user question and a target table containing a target cell corresponding to a target answer for the user question, with the target cell corresponding to a target column and a target row. The computer generates, a first classifier to provide column correlation values reflecting the probability that a given column is the target column. The computer generates a second classifier that provides row correlation values reflecting the probability that a given row is the target row. The computer applies the first classifier to the target table to determine a column correlation value for each column. The computer applies the second classifier to the target table to determine a row correlation value for each row. The computer suggests, as the target cell, a cell having elevated column and row correlation values relative to other target table cells.

A machine learning system for processing image data obtained from an image sensor is provided. The system includes a front end comprising one or more hard-coded filters, each of the one or more hard-coded filters being arranged to perform a set task. The system includes a neural network arranged to receive and process output from the front end. The one or more hard-coded filters include one or more hard-coded noise compensation filters that are hard-coded to compensate for a noise profile of the image sensor from which the image data is obtained. A method of processing image data in a machine learning system is also provided. A system for processing image data is provided.

There are provided a device that collects operation data from equipment, and an information processing apparatus that detects an anomaly or an omen of an anomaly of the equipment on the basis of anomaly detection models constructed from the operation data, the information processing apparatus including means for collecting the operation data, means for learning anomaly detection models from the operation data, and means for calculating an anomaly score of respective operation data from the operation data and the anomaly detection models, the means for learning an anomaly detection model in which a dispersion of elements is small among the anomaly detection models. Thus, in an anomaly detection system, when the operation state of equipment is monitored, even if data for performing division of operation states cannot be obtained or even if division cannot be performed correctly, misdetection of an anomaly such as a malfunction or a failure or an omen of an anomaly can be decreased and the state of the system can be evaluated correctly.

A method for the calculation resource-saving semantic segmentation of image data of an imaging sensor with an artificial neural network, in particular, of a convolutional neural network, the artificial neural network including an encoder path, a decoder path (and a skip component), including: initial connection (merge) of an input tensor to a skip tensor with an initial connection (merge) function/connection instruction to obtain a merged tensor, the input tensor and the skip tensor being dependent on the image data; application of a function of a neural network, in particular, of a convolution to the merged tensor to obtain a proof reader tensor; second connection (merge) of the proof reader tensor to the input tensor with a second connection (merge) function/connection instruction to obtain an output tensor; outputting the output tensor to the decoder path of the artificial neural network.

The present disclosure is directed to a traffic light recognition system and method for advanced driver assistance systems (ADAS) and robust to variations in illumination, partial occlusion, climate, shape and angle at which traffic light is viewed. The solution performs a real time recognition of traffic light by detecting the region of interest, where extracting the region of interest is achieved by projecting the sequence of frames into a kernel space, binarizing the linearly separated sequence of frames, identifying and classifying the region of interest as a candidate representative of traffic light. With the aforesaid combination of techniques used, traffic light can be conveniently recognized from amidst closely similar appearing objects such as vehicle headlights, tail or rear lights, lamp posts, reflections, street lights etc. with enhanced accuracy in real time.

In one embodiment, a method of machine learning and/or image processing for spectral object classification is described. In another embodiment, a device is described for using spectral object classification. Other embodiments are likewise described.

An avatar facial expression generating system and a method of avatar facial expression generation are provided. In the method, multiple user data are obtained and related to the sensing result of a user from multiple data sources. Multiple first emotion decisions are determined, respectively, based on each user data. Whether an emotion collision occurs among the first emotion decisions is determined. The emotion collision is related that the corresponding emotion groups of the first emotion decisions are not matched with each other. A second emotion decision is determined from one or more emotion groups according to the determining result of the emotion collision. The first or second emotion decision is related to one emotion group. A facial expression of an avatar is generated based on the second emotion decision. Accordingly, a proper facial expression of the avatar could be presented.

Method for the calculation resource-saving semantic segmentation of image data of an imaging sensor with the aid of an artificial neural network, in particular, of a convolutional neural network, the artificial neural network including an encoder path, a decoder path, the encoder path transitioning into the decoder path, the transition taking place via a discriminative path, the following steps taking place in the discriminative path: dividing an input tensor as a function of a division function into at least one first slice tensor and at least one second slice tensor, the input tensor originating from the encoder path; connecting the at least one first slice tensor to the at least one second slice tensor as a function of a connection function in order to obtain a class tensor; and outputting the class tensor to the decoder path of the neural network.

A method for brand recognition in video by implementing a brand recognition application coupled to a streaming media player, for identifying an observed set of brands streamed in a broadcast video; receiving, by the brand recognition application, a broadcast video stream of a series of images contained in consecutive frames about an object of interest; extracting a set of brand features from each of image received by applying a trained brand recognition model with neural networks in order to detect one or more features related to each displayed object of interest in each frame, wherein the object of interest is associated with a brand image contained video content displayed to a viewer; and displaying, by a graphic user interface, information from the brand recognition application comprising at least time detected of the brand image in the video content of the broadcast video.