references

An Intuitive Explanation of Connectionist Temporal Classification
chinese_ocr

use

edit eval.py at line 15,replace with your image path

cd cnn-ctc-text-recognition
python3 -m eval.py

demo

['当成为生态文学的主要']

['了采用Google搜']

tensorboard

network detail

Overall stride of cnn network is 8 pixels in width,32 pixels in height