基于前后端分离技术的图像数据集标注系统构建

摘要

The image dataset and its annotation information are the key to the model training of the computer vision system. The traditional data annotation system is mostly based on the B/S architecture. The front-end system has high coupling degree and is not suitable for the crowdsourcing mode application. Therefore, this paper proposes an image dataset labeling system suitable for crowdsourcing by using microservice architecture design technology and using front-end separation. The key technologies such as front-end MVC mode, back-end micro-service architecture, and front-end data interaction constraints in the system design are elaborated.%图像数据集及其标注信息是计算机视系统进行模型训练时的关键,传统的数据标注系统多基于B/S架构,前后端系统耦合度高,不适于众包模式的应用.因此本文利用微服务架构设计技术,采用前后端分离的方式,提出了一种适合众包的图像数据集标注系统.并详细阐述了系统设计中的前端MVC模式、后端微服务架构、以及前后端数据交互约束等关键技术.