Disclosed is a method for performing image convolution by considering a hierarchical relationship of hyperbolic feature vectors in a hyperbolic space. According to an embodiment of the present disclosure, an image convolution method in a hyperbolic space includes steps of embedding an image feature vector on a Euclidean space into a hyperbolic feature vector on a hyperbolic space, allocating a hierarchical weight on the hyperbolic feature vector based on a hierarchical property of the hyperbolic feature vector, and convolutioning the hyperbolic feature vector by applying the hierarchical weight.

Example systems perform complex optimization tasks with improved efficiency via neural meta-optimization of experts. In particular, provided is a machine learning framework in which a meta-optimization neural network can learn to fuse a collection of experts to provide a predicted solution. Specifically, the meta-optimization neural network can learn to predict the output of a complex optimization process which optimizes over outputs from the collection of experts to produce an optimized output. In such fashion, the meta-optimization neural network can, after training, be used in place of the complex optimization process to produce a synthesized solution from the experts, leading to orders of magnitude faster and computationally more efficient prediction or problem solution.

This disclosure describes systems and methods for training a neural network with a training data set including data items labeled at different granularities. During training, each item within the training data set can be fed through the neural network. For items with labels of a higher granularity, weights of the network can be adjusted based on a comparison between the output of the network and the label of the item. For items with labels of a lower granularity, an output of the network can be fed through a conversion function that convers the output from the higher granularity to the lower granularity. The weights of the network can then be adjusted based on a comparison between the converted output and the label of the item.

Embodiments provide systems and techniques for automated haptics generation and distribution. An example technique includes receiving media content from a computing device. The media content includes at least one of audio content or video content. One or more features of the media content is determined. A set of haptic data is generated for the media content, based on evaluating the one or more features of the media content with at least one machine learning model. Another example technique includes obtaining a set of haptic data associated with media content. The set of haptic data, metadata, and the media content is transmitted to a computing device.

The present disclosure provides an image processing method, an image processing device, an electronic apparatus and a readable storage medium. The image processing method includes: obtaining feature map data of an input image; extracting a feature region in the feature map data in accordance with a size of a convolution kernel; performing windowing processing on the feature region; and obtaining a windowed feature map of the input image in accordance with the feature region obtained after the windowing processing.

In a foreground extraction apparatus, an extraction result generation unit performs a foreground extraction using a plurality of foreground extraction models for an input image, and generates foreground extraction results. A selection unit selects one or more foreground extraction models among the plurality of foreground extraction models using respective foreground results acquired by the plurality of foreground extraction models. A foreground region generation unit extracts each foreground region based on the input image using the selected one or more foreground extraction models.

A feature learning system (100) includes a similarity definition unit (101), a learning data generation unit (102), and a learning unit (103). The similarity definition unit (101) defines a degree of similarity between two classes related to two feature vectors, respectively. The learning data generation unit (102) acquires the degree of similarity, based on a combination of classes to which a plurality of feature vectors acquired as processing targets belong, respectively, and generates learning data including the plurality of feature vectors and the degree of similarity. The learning unit (103) performs machine learning using the learning data.

A computer-implemented clustering method is disclosed for image segmentation, social network analysis, computational biology, market research, search engine and other applications. At the heart of the method is a point-set kernel that measures the similarity between a data point and a set of data points. The method has a procedure that employs the point-set kernel to expand from a seed point to a cluster; and finally identifies all clusters in the given dataset. Applying the method for image segmentation, it identifies several segments in the image, where points in each segment have high similarity: but points in one segment have low similarity with respect to other segments. The method is both effective and efficient that enables it to deal with large scale datasets. In contrast, existing clustering methods are either efficient or effective; and even efficient ones have difficulty dealing with large scale datasets without massive parallelization.

A computer-implemented clustering method is disclosed for image segmentation, social network analysis, computational biology, market research, search engine and other applications. At the heart of the method is a point-set kernel that measures the similarity between a data point and a set of data points. The method has a procedure that employs the point-set kernel to expand from a seed point to a cluster; and finally identifies all clusters in the given dataset. Applying the method for image segmentation, it identifies several segments in the image, where points in each segment have high similarity: but points in one segment have low similarity with respect to other segments. The method is both effective and efficient that enables it to deal with large scale datasets. In contrast, existing clustering methods are either efficient or effective; and even efficient ones have difficulty dealing with large scale datasets without massive parallelization.

An urban remote sensing image scene classification method in consideration of spatial relationships is provided and includes following steps of: cutting a remote sensing image into sub-images in an even and non-overlapping manner; performing a visual information coding on each of the sub-images to obtain a feature image Fv; inputting the feature image Fv into a crossing transfer unit to obtain hierarchical spatial characteristics; performing convolution of dimensionality reduction on the hierarchical spatial characteristics to obtain dimensionality-reduced hierarchical spatial characteristics; and performing a softmax model based classification on the dimensionality-reduced hierarchical spatial characteristics to obtain a classification result. The method comprehensively considers the role of two kinds of spatial relationships being regional spatial relationship and long-range spatial relationship in classification, and designs three paths in a crossing transfer unit for relationships fusion, thereby obtaining a better urban remote sensing image scene classification result.