Method and system for active data labeling based on maximum information triple screening network

Abstract

The invention relates to the field of data labeling, in particular to a method and system for actively labeling data based on the largest information triple screening network, with the purpose of reducing the workload of manual labeling while ensuring labeling accuracy. According to the depth characteristics of the samples, the present invention selects the most valuable unlabeled samples for manual labeling, and constructs the largest information triplet loss function; gradually updates the data structure and network parameters, and then updates the screening network model. Stop updating the screening network model until the maximum intra-class differences between the unlabeled data and the labeled data are less than the preset second threshold and the minimum inter-class differences are greater than the preset first threshold. The remaining data can be annotated by computer using the last updated screening network model. Through this method, the accuracy of labeling is ensured while reducing the workload of manual labeling.

Description

technical field [0001] The invention relates to the field of data labeling, in particular to a method and system for actively labeling data based on the largest information triplet screening network. Background technique [0002] With the advent of the big data era and the development of hardware technology, the research on large-scale databases has been continuously innovated, and at the same time, we have seen the unstoppable advantages of deep learning in computer vision. [0003] But the dominance of deep learning is inseparable from the support of massive labeled data. With the continuous expansion of data volume, manual labeling of massive data is not only costly, but also time-consuming and labor-intensive. Therefore, the annotation of massive data has always been a concern in the field of image annotation. Although deep learning technology has achieved remarkable success in computer vision, due to the small amount of labeled data in the image label itself, deep lea...

AI Technical Summary

Problems solved by technology

This method generally only considers the high uncertainty of the labeled samples between classes, and the process of calculating the information entropy of each sample one by one and then evaluating is too cumbersome

Although some methods use screening representative samples for manual labeling, they do not consider intra-class differences when screening representative samples.

In addition, the optimization of the classifier does not take full advantage of the filtered data samples

Images