method: Attention based method for scene text recognition2019-05-01

Authors: someone

Description: This is an end-to-end framework for arbitrary scene text recognition, which contains a CNN block to extract the feature from the input image, a RNN-based Encoder and a RNN-based Decoder. There is a attention block to locate the position of character in feature map in Decoder. In test stage, the decodes will output one character at each time step until the END token appear.