Authors: 刘杰，杜鹏程

Description: densenet base model with ctc loss , pretrain model train with VGG Synthetic Word Dataset and Synthetic Chinese String Dataset.