2 comments

[ 3.1 ms ] story [ 14.4 ms ] thread
Very cool. Now hurry up and release it.
Interesting article! I had a similar idea in one of my earlier captcha cracking approaches. For me, it had several problems:

1) it was very hard to get the captcha generation exactly right, so that it is close enough to the algorithm of the original site

2) comparing two generated images containing text meaningfully (in a noise tolarant way) is hard

Eventually I went with more regular computer vision techniques.

It's nice to see that the approach is feasible with the right techniques :)