How to make the machine to understand each style and font is different number is a problem

For Google, which uses a street-view car to gather map information, it is a question of how to make the machine understand that each style and font is different, thus deriving a Google image recognition algorithm--according to Google's updated monitors on ICL in recent days, Google's recognition accuracy in this area has reached more than 90%.

But it also poses a problem: if machines can easily identify images that contain text messages, then the verification code for security on each site can be easily deciphered, Google said, and they found that there is an algorithm at present that can make deciphering more than 99.8% accurate.

To do that, Google has to deal with it-by distorting the font and refactoring the code so that its reCAPTCHA project can cope with the decoding algorithm by studying how it is difficult for the machine to understand and how humans can understand it. If you are interested, you can click to view the reCAPTCHA project and Google's latest updated research paper.