Image set establishing method and device and storage medium

Abstract

The embodiment of the invention discloses an image set establishing method and device and a storage medium. The method comprises the steps of obtaining a single-label image set and a multi-label imageset; converting the contents of the tags into word identifiers according to a semantic network to obtain a word identifier set, a converted single-tag image set and a converted multi-tag image set; constructing a hierarchical semantic structure according to the word identifier set and the semantic network; carrying out label supplement on the images in the converted single-label image set according to the semantic relationship between the word identifiers in the hierarchical semantic structure to obtain a supplemented single-label image set; based on the co-occurrence probability between theword identifier in the supplemented single-label image set and the word identifier in the converted multi-label image set, carrying out label supplementing on the images in the supplemented single-label image set to obtain a final supplemented image set; and establishing a target multi-label image set according to the finally supplemented image set and the converted multi-label image set. According to the scheme, the establishment efficiency of the large-scale multi-label image set and the label labeling quality can be improved.

Description

technical field [0001] The invention relates to the technical field of communication, in particular to a method, device and storage medium for establishing an image set. Background technique [0002] With the development of deep learning models and training methods, great progress has been made in the field of computer vision, and the research direction has gradually shifted from low-level image processing and image recognition to higher-level visual understanding. Complex vision tasks require the utilization of deep neural networks with better visual representation potential, and an important condition for converting this potential into visual representation capabilities is large-scale image datasets. [0003] Among them, the establishment of large-scale image data sets mainly involves image sources and image annotations; at present, image sources are very simple, and a large number of images can be obtained cheaply from the Internet, but the cost of image annotation is ver...

AI Technical Summary

Problems solved by technology

[0005] At present, the establishment of large-scale multi-label image sets still adopts the traditional manual labeling method, that is, manually labeling each image with multiple labels; due to the slower labeling speed of manual labeling methods, the efficiency of large-scale multi-label image set establishment is low.

Images