Annotation quality detection device and method

Abstract

The invention discloses an annotation quality detection device, and the device comprises a before-annotation and after-annotation consistency detection module which is used for detecting the consistency of annotation data formed by annotating the same to-be-annotated object data in different time periods before and after the same class of annotation tasks in the annotation process; an annotation collaborative consistency detection module which is used for detecting the consistency of annotation data formed by annotating the same to-be-annotated object data by different types of annotation tasks in the annotation process; an annotation accuracy detection module which is used for taking annotated data of the to-be-annotated object data as reserved annotation data in the annotation process, annotating the to-be-annotated object data to form corresponding annotation data, and performing accuracy comparison on the annotation data and the reserved annotation data to form third annotation quality; and a quality comprehensive module which is used for determining comprehensive annotation quality according to the three annotation qualities. The invention further provides an annotation quality detection method. According to the invention, the annotation quality can be automatically detected.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a marking quality detection device. The invention also relates to a marking quality detection method. Background technique [0002] In the field of artificial intelligence (AI), it is often necessary to label data to realize data recognition, so the quality of data labeling is very important. When the quality of labeling is poor, there may be recognition deviation or failure to identify situation. [0003] In the existing technology, the amount of data and workload involved in data labeling are huge. At present, for large-scale data labeling, there is no good tool for testing the quality of labeling data, and quality testing can only be performed manually. For different labeling tasks, such as full check or spot check, the methods that may be used are also different. [0004] Those skilled in the art should know that manual inspection is not only inefficient, but also the inspec...

AI Technical Summary

Problems solved by technology

[0004] Those skilled in the art should know that when using manual inspection, not only is the efficiency low, but also the inspectors will inevitably make mistakes when the workload is heavy, so there is an urgent need for a method that can realize automatic labeling quality inspection

Images