CNN-CTC recognizer

[sanakhamekhem]

Dear Sir,
I'm intersted to this code. I'm building a recognizer for images having a height of 128 pixels.
How can I update just the CNN recognier architecture's model to be suitable with my input images.
Thank you in advance.