Skip to content

teddyjkwang/STN_CNN_LSTM_CTC_TensorFlow

This branch is 5 commits behind wushilian/STN_CNN_LSTM_CTC_TensorFlow:master.

Folders and files

NameName
Last commit message
Last commit date

Latest commit

b07e996 · Jan 19, 2018

History

38 Commits
Dec 17, 2017
Nov 10, 2017
Dec 17, 2017
Nov 5, 2017
Nov 30, 2017
Jan 19, 2018
Nov 14, 2017
Nov 30, 2017
Nov 30, 2017

Repository files navigation

STN_CNN_LSTM_CTC_TensorFlow

use STN+CNN+BLSTM+CTC to do OCR you can choose the basecnn(CRNN) or densenet

Attention

It's hard to converge use STN,so you can delete the STN in model ,and it's easy for you.If you delete the STN,the number of val image don't need equal to 256

How to use this model

1.put your train iamge to 'train' dir,and image name should be like index_label_.jpg(1_abc_.jpg)

2.put your val image to 'val256' dir,and the number of val image should be equal to batch_size(default 256)

3.run train.py

Use STN in mnist

left is rotated image,right is transformed image(STN's output)

image

About

use STN+CNN+BLSTM+CTC to do OCR

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%