"The SCUT-CTW1500 dataset contains 1,500 images, with 1,000 used for training and 500 used for testing. Specifically, it provides 10,751 text instance images, of which 3,530 contain curved text. These
Total-Text is a text detection dataset that consists of 1,555 images with a variety of text types including horizontal, multi-oriented, and curved text instances. The training split and testing split