Context Everytime I read a paper or anything abut image processing the examples are always easy and most of time don't reflect the real world with real problems. So I decided to create a dataset with captchas because if it's possible to OCR a captcha we can OCR anything Content The dataset contains 300 captcha images already solved (Train and test) and some not solved yet (non identified). To help users and improve the discussions about image processing and technics the folders named treated contains the same images of train folder but with some image processing All the treatments done are in the kernel.