Authors: Hui Tan, JieDong Hao, YaFei Wen
Description: Our method is based mainly on CNN + BLSTM + CTC. Both horizontal and vertical model are used, and a language model is applied to the final phase for correcting the text predicted.
Synthetic text data generated by our synthtext-tool, and real scene text data including public dataset such as RCTW17, LSVT, ReCTS, CASIA-10K, ICPR18, are used for training.
Hui Tan, JieDong Hao, YaFei Wen
vivo AI Lab
Shi B, Bai X, Yao C. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(11): 2298-2304.