Picture annotation data obtaining method and system

Abstract

The invention provides a picture annotation data obtaining method and system. The method comprises the steps of obtaining an identification result of a to-be-annotated picture; displaying the to-be-annotated picture and candidate characters in a word guessing interface, wherein the candidate characters at least include characters obtained by splitting the identification result; and by utilizing achoice of a user to the candidate characters in the word guessing interface, obtaining annotation data of the to-be-annotated picture. According to the picture annotation data obtaining method and system, the to-be-annotated picture is presented to the user in a word guessing game form, and the user finishes the annotation of the picture through word guessing; a large amount of computing resourcesdo not need to be occupied, and professional annotation or outsourcing personnel do not need to be recruited, so that the picture annotation data obtaining cost is reduced; and the user plays the game according to self will, and each calibrated task can obtain point awards, so that the achievability and positivity of playing the game by the user are improved, the cheat possibility is basically eliminated, and the accuracy is improved to a certain extent.

Description

【Technical field】 [0001] The present application relates to the field of computer processing technology, and in particular to a method and system for acquiring picture annotation data. 【Background technique】 [0002] Among the massive information generated and stored on the Internet, pictures are an important information carrier. In Internet information provision and information retrieval services, the processing of picture information is becoming more and more important. [0003] Image annotation is a very important task in the preparation of training data in the field of computer vision. It usually requires a large number of manually annotated images as the initial training data set for further data processing and data mining of machine learning and computer vision. Existing methods for obtaining manual annotation data for images mainly manually annotate images through specialized annotators, outsourcing or crowdsourcing. [0004] However, since image labeling is a boring...

AI Technical Summary

Problems solved by technology

[0004] However, since image labeling is a boring, simple and repetitive work, long-term labeling will cause physical and psychological harm to the labelers, and the labor cost is high

[0005] In addition, when the number of pictures that need to be manually labeled is large, there is also the possibility of cheating. Some labelers may randomly label some pictures in the process of labeling in order to complete the labeling task quota as soon as possible.

Images