Pycapt is a series of image processing algorithms I wrote while processing the captcha. You can use it to denoise, disturb lines and split the captcha. pycapt encapsulates some methods about the graph matrix, such as splitting the image and croping the imgs.
Standardized matrices, generating the training set images you need, and more can help you use deep learning for image recognition.
GitHub welcomes pr, if there are bugs or new requirements, please feedback issue.