Class identification method for zero sample picture

Abstract

The invention discloses a class identification method for a zero sample picture. The method includes constructing a knowledge graph according to the knowledge of the human to represent explicit relationships among the categories; according to the method and the device, the problems of implicit relation learning and fuzzy and uncertain relation between categories in a semantic space are avoided, meanwhile, a residual image convolutional network is constructed and trained for migrating knowledge between the categories, and category recognition is carried out by adopting maximum inner product values, so that the category recognition accuracy of the zero sample picture is improved.

Description

technical field [0001] The invention belongs to the technical field of picture recognition, and more specifically, relates to a category recognition method of a zero-sample picture. Background technique [0002] Traditional image category recognition methods require a large amount of labeled data for training, and the trained classifier can only identify the categories of pictures participating in the training, and cannot do anything for new categories of pictures. However, the process of annotating images is time-consuming and expensive, and it is difficult to obtain a large number of annotated samples in reality. [0003] The goal of category recognition from zero-shot images is to identify the categories of images that do not appear in the training set. The existing zero-shot image category recognition methods are mainly divided into two types. [0004] The first type is to learn a common semantic space in which category knowledge acquired on the training set is transfe...

AI Technical Summary

Problems solved by technology

[0007](1), attribute labeling is expensive and time-consuming, and attribute-based methods are not suitable for category recognition of large-scale zero-sample images;

[0008] (2), the word vector information is learned from a large corpus, which has a large error, and the relationship between categories is usually learned through word vector information Inaccurate;

[0009] (3), learn the implicit relationship between categories with the help of semantic space, however, the relationship between categories is fuzzy and uncertain, at the same time, It is very difficult to learn the implicit relationship between categories in the semantic space, resulting in low recognition accuracy of zero-shot images

Images