Total-Text
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/cs-chan/total-text-dataset
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了1255张训练图像和300张测试图像，每张图像中至少有一个曲线样本，并使用多边形边界框进行标注。此外，该数据集专为测试模型在曲线文本实例上的表现而设计。规模上，总共有1555张图像。其任务定位为文本识别。
This dataset comprises 1255 training images and 300 test images, resulting in a total of 1555 images overall. Each image contains at least one curved text sample, and all samples are annotated with polygonal bounding boxes. Additionally, this dataset is specifically developed to evaluate model performance on curved text instances, with its core task being text recognition.
提供机构：
Total-Text
