Self-paced cross-modal matching method based on subspace

Abstract

The invention discloses a self-paced cross-modal matching method based on subspace. The method is characterized by extracting feature vectors of different modals of data in a data set; obtaining different mapping matrixes corresponding to different modals in a training set through subspace self-paced learning, and mapping the different modal types of data samples in a test set to the same space through the mapping matrixes to enable the data in the training set and the test set to be mapped to a unified space; and then, measuring the similarity between search data and target data in the test set to obtain a cross-modal matching result The method can enable the different modals of data to be mapped to the unified space, and sample selection and feature learning are carried out while mapping, thereby improving matching robustness and accuracy.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a subspace-based self-stepping pattern matching method. Background technique [0002] Data in reality often have multiple modalities. For example, web page data includes both picture information and text information; video data includes audio information and picture information at the same time. The fundamental task of cross-modal matching is to use a modality as a query condition to match similar heterogeneous modality information. Most of the traditional cross-modal matching methods are supervised learning methods. They all use semantic labels to reduce the gap between heterogeneous modalities, but they cannot deal with unlabeled information, and manually labeling data is a time-consuming and laborious work. In addition, some unsupervised methods do not consider the judgment and correlation of features and the semantic similarity between samples, which cannot meet...

AI Technical Summary

Problems solved by technology

Most of the traditional cross-modal matching methods are supervised learning methods. They all use semantic labels to reduce the gap between heterogeneous modalities, but they cannot deal with unlabeled information. Manually labeling data is a time-consuming and laborious work.

In addition, some unsupervised methods do not consider the judgment and correlation of features and the semantic similarity between samples, which cannot meet people's daily needs

Images