RCTW-17 is a competition for reading Chinese text images. Its full name is ICDAR2017 Competition on Reading Chinese Text in the Wild. The RCTW-17 dataset is the basis of this competition.

The dataset contains more than 12,000 images, mainly for text detection and text recognition, most of which are collected in the wild through mobile phone cameras, and some are screenshots.

The RCTW-17 dataset was released by the Vision and Deep Learning Research Group of Huazhong University in April 2017. The main publishers were Baoguang Wang, Cong Yao, Serge Belongie, Shijian Lu and Xiang Bai.