Abstract: The semantic segmentation data used in training model of image segmentation algorithm can only be used after annotation. At present, the open source data annotation tools related to semantic segmentation only support the annotation of object contour by marking polygons by tracing points, which has the disadvantages of many operations and low annotation efficiency. In order to solve the problems, this paper proposes the mouse fast annotation method and the whole copy/adjust annotation polygon method, which are applied to the data annotation system. The experimental results show that the annotation efficiency can be improved by 11.6% and 29.7% respectively by using the mouse fast annotation method and the whole copy/adjust annotation polygon method. Using this data annotation system for image segmentation can effectively improve the annotation efficiency of the annotator and have a positive effect on the acquisition of a large number of high quality annotation data.

Key words: data annotation, polygon annotation, annotation system

摘要： 图像分割算法训练模型用到的语义分割数据需要经过标注才能使用，目前语义分割相关的开源数据标注工具仅支持描点标注多边形方式标注物体轮廓，存在操作繁多、标注效率低下的缺点。针对上述问题，提出鼠标快速标注方法和整体拷贝/调整标注多边形方法，并应用于数据标注系统。实验结果表明，使用鼠标快速标注方法和整体拷贝/调整标注多边形方法标注分别可以提高11.6%的标注效率和29.7%的标注效率。使用该面向图像分割的数据标注系统可以有效提高标注人员的标注效率，对大量高质量标注数据的获取有积极作用。

关键词: 数据标注, 多边形标注, 标注系统