Curved Synthetic Dataset 150k
收藏arXiv2025-09-30 收录
下载链接:
https://github.com/Yuliang-Liu/SPTSv2
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150,000张为提升文本检测器性能而生成的合成图像,其中大约三分之一是曲线文本,而剩下的三分之二为水平文本实例。在此数据集上进行预训练可以增强模型在场景文本检测任务上的性能。该数据集的任务是文本检测。
This dataset comprises 150,000 synthetic images generated to enhance the performance of text detectors. Approximately one-third of these images are curved text instances, while the remaining two-thirds are horizontal text instances. Pretraining on this dataset can improve the model's performance on scene text detection tasks. The task of this dataset is text detection.
提供机构:
SynthText toolbox



