#SUST-and-RUST-datasets-for-Uyghur-STR
SUST is a synthetic dataset containing 600,000 Uyghur images, each image contains one Uyghur word, the maximum word length is 15 characters and the word consists of only 32 basic Uyghur alphabets without other symbols. Therefore, this dataset is only suitable for training word-level scene-based Uyghur text recognition models. The synthesis tool used is TextRecognitionDataGenerator. The background of each picture is randomly selected from more than 8000 background materials, the font is randomly selected from more than 300 kinds of Uyghur fonts, and the text in the picture has the effects of tilting, blurring and distorting in random degrees, as shown in Figure. The backgrounds, fonts, and corpus in SUST are sourced from the internet.
RUST is a real dataset shot in Xinjiang, China, as shown in Fig. 5, which contains a total of 4,000 real Uyghur scene text images. as shown in Figure.
