Abstract : In this report we discuss the application of machine learning techniques to the problem of reading text in CAPTCHAs. We first present techniques for reading a CAPTCHA that requires only single character classification and trivial segmentation, and show that we are able to successfully read such CAPTCHAs.

We then show how we attempted to read a more difficult CAPTCHA; one that required significant effort to segment before individual character classification can be attempted. We detail our noise removal, line removal, and clustering-based segmentation techniques, and present our results for this second type of CAPTCHA



Please find the following attachments "Project Report on Captcha Semianr Report/pdf/ppt download"here......