Blue Eye Macro

Hello

So there is this simple captcha..













There is some way to make BEM identify letters patterns?

If anyone know a way to config BEM on solving this simple captcha, i will be very thankful!

Hi,

I have done a simple captcha bot Before so here is what I have done.

1. Identify what's the Captcha want you to do or find in you case click a number or letter, a-z then 0-9.
2. The color of the Characters you need to find in your case the color is all the same. (however in the choices it is different but the color in the first , second, 3rd and 4th are constant if even if you are to find the different color so all you need to do is change the color of the correpoding color per choices in the patter that you will make)
3. The Orientation of the Characters you need to find in your case they are twisted or tilt in random angle Making the pattern Variable so you may need 2-4 pattern per character(this is not impossible but a lot of work)
4. The size of the characters if they are the same of they have different sizes(if they are then you need pattern for each size)
so if there are 4 different orientation and 3 different sizes you will need to make 12 pixel pattern per character.
multiply by (26+10= 30). 30 x 12 = 360 pattern all in all.
5. Start Getting the patterns then Try it if the BEM can find it.
6. Once you have the pixel Patterns start creating the code
first limit the area of interest to the area of the character you need to find
7. then if pixel pattern of character "a" can be found.
begin
unlimit area of interest
limit area of of interest to the first character in the choices
if pixel pattern of character "a" can be found(remember to change RGB to the corresponding color of the Choices)
begin
mouse click at the coordinates of the first choice.
end
if pixel pattern of character "b" can be found
begin
mouse click at the coordinates of the first choice.
end
.
.(do the rest)
.
.
if pixel pattern of character "9" can be found
begin
mouse click at the coordinates of the first choice.
end
unlimit area of interest
limit area of of interest to the first character in the choices
if pixel pattern of character "a" can be found
begin
mouse click at the coordinates of the first choice.
end
if pixel pattern of character "b" can be found
begin
mouse click at the coordinates of the first choice.
end
.
.(do the rest)
.
.
if pixel pattern of character "9" can be found
begin
mouse click at the coordinates of the first choice.
end
unlimit area of interest
limit area of of interest to the first character in the choices
if pixel pattern of character "a" can be found(remember to change RGB to the corresponding color of the Choices)
begin
mouse click at the coordinates of the first choice.
end
if pixel pattern of character "b" can be found
begin
mouse click at the coordinates of the first choice.
end
.
.(do the rest)
.
.
if pixel pattern of character "9" can be found
begin
mouse click at the coordinates of the first choice.
end
unlimit area of interest
limit area of of interest to the 2nd character in the choices
if pixel pattern of character "a" can be found
begin
mouse click at the coordinates of the first choice.
end
if pixel pattern of character "b" can be found
begin
mouse click at the coordinates of the 2nd choice.
end
.
.(do the rest)
.
.
if pixel pattern of character "9" can be found
begin
mouse click at the coordinates of the 2nd choice.
end
.
.(do the rest till the 4th choice)
.
end
unlimit area of interest
limit area of of interest to the first character in the choices again
if pixel pattern of character "b" can be found.
begin
.
.repeat till you reach the last pixel pattern in the character you need to find( I think it would be better to use function)
.
end
end

something like that you can do the rest and find which is better if begin loop() or Begin loop(2)
or use break from loop.
hope this give some insight.

It will be a very long code. Function would help to shorter it I guess,

This captcha can be easily be solved via neural network.
Recognizing numbers (and letters) is the very first example in every deep learning handbook.

When you're able to recognize character, you can do it for the one on the left and for the ones on the right, and find a match.

If you want to learn more about neural network without having any pre-knowledge, I would suggest you this video:
https://www.youtube.com/watch?v=aircAruvnKk

At the moment there is no easy way to implement a neural network on BEM, but I'm gonna work on it (I hope) in the next months.