Authors: 杜鹏程

Description: We use ctpn and densenet+ctc loss to solve this problem.The pretrained model that we use is based on SynthText_Chinese_version data training.