摘要
图像已成为数字人文研究新的热点,而图像标注是图像知识传递和价值揭示的主要途径.目标检测作为一种计算机视觉和图像处理领域的技术,可快速批量实现图像中目标对象的定位与分类.基于目标检测相关原理,本次尝试是将视觉目标标注工具VoTT和多维度图像智慧系统MISS结合,形成完整的图像应用生态流程.同时提出了图像多层标注模型框架——包含图像层、对象层(内容和标签)、语义层三层四个维度,其中标注区域的选定和对象层的标签使用了VoTT自动标注的结果.最后通过案例对所提框架和技术进行了论证,以期为数字人文图像研究提供更多的应用视角.
Image has become a new hot spot in digital humanities research,and image annotation is the main way of image knowledge transfer and value reveal.Object detection,as a computer technology related to computer vision and image processing,can quickly realize the location and classification of target objects in images in batches.The paper first introduces the relevant principles of object detection,and proposes a complete image application ecological process which combines the Visual Object Tagging Tool(VoTT)and the Multi-Dimensional Image Smart System(MISS).At the same time,an image multi-layer annotation model framework is proposed which contains four dimensions:image layer,object layer(content and label),and semantic layer.The selection of the annotation area and the label of the object layer will use the results of automatic labeling with VoTT.In the end,the paper exemplifies the proposed framework and technologies through cases,in order to provide more application perspectives for the research of digital humanistic images.
出处
《数字人文研究》
2021年第3期39-50,共12页
Digital Humanities Research
基金
国家社会科学基金一般项目“数字人文中图像文本资源的语义化建设与开放图谱构建研究”(编号:19BTQ024)研究成果之一
关键词
目标检测
数字人文
图像标注
国际图像互操作框架
object detection
digital humanities
image annotation
International Image Interoperability Framework(IIIF)
作者简介
陈涛,中山大学信息管理学院副教授,Email:chent283@mail.sysu.edu.cn.
