SCUT-CTW1500
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/yuliang-liu/curve-text-detector
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个包含1000张训练图像和500张测试图像的基准，这些图像在文本行级别使用多边形进行标注，特别关注任意形状的场景文本。此外，该数据集是专门为识别任意形状的场景文本而设计的。规模上，总共有1500张图像。其任务是对场景文本进行检测。
This dataset is a benchmark consisting of 1,000 training images and 500 test images. All images are annotated with polygonal labels at the text-line level, with a particular focus on arbitrary-shaped scene text. Moreover, this dataset is specifically designed for arbitrary-shaped scene text recognition. In total, the dataset contains 1,500 images. The core task of this benchmark is scene text detection.
提供机构：
SCUT
