Method and device for training a multi-label classification model

Abstract

The invention provides a method and a device for training a multi-label classification model, which can dynamically learn image features, enable a feature extraction network to better adapt to task requirements, and are good in multi-label classification effect. The method comprises the steps of determining n samples and a label matrix Yc * n corresponding to the n samplesfrom a training data set,wherein an element yi * j in the label matrix Yc * n represents whether an ith sample contains an object indicated by a jth label or not, and c represents the number of labels related to the samples;extracting a feature matrix Xd * n of the n samples by using a feature extraction network; using a first mapping network to obtain a prediction label matrix of the characteristic matrix Xd * n, and using a second mapping network to obtain a low-rank label matrix of the label matrix, wherein the prediction tag matrix and the low-rank tag matrix are defined in the specification; updating the weightparameter Z, the feature mapping matrix Mc * d and the low-rank label correlation matrix S, and training the multi-label classification model.

Description

technical field [0001] The present application relates to the computer field, and more specifically, to a method and device for training a multi-label classification model in the computer field. Background technique [0002] With the improvement of the processing performance of smart phones, more and more applications have put forward requirements for image recognition. For example, in the process of taking pictures with a mobile phone, if the smartphone can accurately identify objects within the shooting range, it can perform targeted calculations on its color and shape, thereby improving the shooting effect. In the machine learning of intelligent systems, the training of recognizing objects in images has become a very important aspect. Generally speaking, machine learning is to set labels for a large number of existing images for the objects contained in them, and then continuously adjust the recognition parameters through computer self-evolution to gradually improve the ...

AI Technical Summary

Problems solved by technology

After the features of the image are extracted, the features of the image are fixed, so it is not possible to dynamically learn the feature information of the input image according to the label.

Images