Total-Text
收藏资源简介:
Total-Text数据集由马来亚大学图像与信号处理中心创建,专注于场景文本检测与识别,特别是曲线路径文本。该数据集包含1555张图像,共9330个标注文本实例,涵盖水平、多向和曲线文本三种方向。创建过程中,数据集考虑了文本方向的多样性,超过半数图像包含两种以上文本方向。Total-Text旨在解决现有数据集中曲线文本缺失的问题,推动场景文本检测技术的发展,特别是在处理复杂场景中的曲线文本识别方面。
Total-Text dataset was created by the Centre for Image and Signal Processing at the University of Malaya, focusing on scene text detection and recognition, particularly curved text. This dataset includes 1555 images with a total of 9330 annotated text instances, covering three text orientations: horizontal, multi-oriented, and curved text. During its construction, the diversity of text orientations was fully considered, with over half of the images containing more than two types of text orientations. The core goal of Total-Text is to address the lack of curved text samples in existing datasets, and promote the development of scene text detection technologies, especially for curved text recognition in complex scenes.




