1
$\begingroup$

I'm trying to complete a captcha, and here is what it looks like:

enter image description here

Between captchas the calligraphy of the letters is the same, but the letters may be resized and rotated. And the background noise (the small dots and lines over and around the letters) will be different. Any Hangul letter may appear.

Edit 1: I can generate any number of new captchas with an answer for each of them. But to be clear, the answers that are generated are for entire captchas, that is, multiple Hangul letters arranged in a specific order as the answer for each captcha, not for individual letters.

  1. What type of machine learning is best for this problem?
  2. How do I extract good data from the image above for this problem?

Update 1: Unfortunately no one has given any suggestions for how to solve this yet. My idea at the moment is to mimic the model in this paper: https://www.ics.uci.edu/~xhx/publications/HHR.pdf

$\endgroup$
2
  • $\begingroup$ Do you have access to the original Captcha generator? That is the best source for training data, as it will generate pairs of images and correct answers. If you don't have access, could you please clarify "academic purposes"? For most academic purposes, it will not matter which precise generator you use, so an answer would point you to use a different generator that you do have access to - you would learn the same things by creating an image processor that could crack a different Captcha . . . $\endgroup$ – Neil Slater Jun 13 at 17:33
  • $\begingroup$ Thank you for your reply. I do have access to the captcha generator and I can generate any number of new captchas with an answer for each of them. But to be clear, the answers that are generated are for entire captchas, that is multiple hangul letters arranged in a specific order as the answer for each captcha, not for individual letters. I edited my post to include this information. $\endgroup$ – I take people's shoes Jun 13 at 19:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.