BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and

BaiXiang的CRNN论文阅读
1.论文题目
BaiXiang——【arXiv2015】An End-to-End Trainable Neural Network forImage-based Sequence Recognition and Its Application to Scene Text Recognition
2.论文思路和方法
1)问题范围: 单词识别
2)CNN层:使用标准CNN提取图像特征,利用Map-to-Sequence表示成特征向量;
3)RNN层:使用双向LSTM识别特征向量,得到每列特征的概率分布;
4)Transcription层:利用CTC和前向后向算法求解最优的label序列;
3.亮点和创新点
1)端到端可训练(把CNN和RNN联合训练)
2)任意长度的输入(图像宽度任意,单词长度任意)
【BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and】3)训练集无需有字符的标定
4)带字典和不带字典的库(样本)都可以使用
5)性能好,而且模型小(参数少)
4.相关链接
1)白翔的个人主页:http://mc.eistar.net/~xbai/
2)论文的下载地址:https://arxiv.org/pdf/1507.05717v1.pdf
3)代码的下载地址:
http://mc.eistar.net/~xbai/CRNN/crnn_code.zip
5.论文细节
1)论文的框架
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片


2)特征提取层
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片


3)序列标定层
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片


4)翻译层
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片


BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片


5)网络训练
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片



6)实验
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片



7)总结
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片



8)问题
BaiXiang——【arXi2015】An End-to-End Trainable Neural Network for Image-based Sequence Recognition and
文章图片





    推荐阅读