【ocr：end to end】ASTER ：An Attentional Scene Text Recognizer with Flexible Rectification

最近从目标检测和分类转到了做OCR，什么都不太懂，只能一点一点的去理解：

文中需要学习的知识点：

Sequence to Sequence Learning with Neural Networks

sequence to sequence model小记

Beam Search Algorithm

理解LSTM（通俗易懂版）

Thin Plate Spline (薄板样条函数)

薄板样条函数(Thin plate splines)的讨论与分析

Position Embedding

什么是Attention

 

文中提到的论文：

1.《An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition》，即CRNN网络；

2.《spatial transformer networks》，解释了什么是STN；

3.《Neural Machine Translation by Jointly Learning to Align and Translate》，能做到focus和这个目标最相关的输入；

4.《Attention-based models for speech recognition》，介绍了Attention在语音识别中的应用；

 

别人对该文章的理解:

ASTER_An Attentional Scene Text Recognizer with Flexible Rectification