SDK Question
As long as i browsed through the forum i understand @Sven is the developer of CB, right? Well, congrats for the great product!
I am using CB and it does very good job for me, however i am hitting captchas not included in the db and i want to train CB for them. For some i managed to train, for others not yet..
What i wanted to ask is about the last "Use Masks" solving method, as i can't seem to understand how it works and there's no info at all about it. I have small bit of programming knowledge and i was able to understand what is doing the "use hash values" and "use fixed results" and how they should be used, but the "Use Masks" method i have no idea about and to learn to use it i need to know what it does and how it should be used properly.
@Sven can you please give me a bit of info about the "Use Masks" method.
For what kind of images it is good for?
Does it work for any multicolor images?
Rotated / distorted characters?
I see it makes some string values for each character like the "use hash values", but those don't look like hashes. Will it work with images having characters at different positions as there there's no box region like in "use hash values", so i assume it auto determines position of characters or sets it as fixed from the sample images?..
In some of your definitions i see you make the images black & white (thresholded) first with the filters, for others not.. so does this method rely on having the characters in exactly same colors like the "use hash values" relies on to make same hash value?
What the "must match atleast by X%" and "Add if match is less than X%" options mean and do?
I will appreciate your input as that will make me understand this solving method and help me make better solving definitions
I am using CB and it does very good job for me, however i am hitting captchas not included in the db and i want to train CB for them. For some i managed to train, for others not yet..
What i wanted to ask is about the last "Use Masks" solving method, as i can't seem to understand how it works and there's no info at all about it. I have small bit of programming knowledge and i was able to understand what is doing the "use hash values" and "use fixed results" and how they should be used, but the "Use Masks" method i have no idea about and to learn to use it i need to know what it does and how it should be used properly.
@Sven can you please give me a bit of info about the "Use Masks" method.
For what kind of images it is good for?
Does it work for any multicolor images?
Rotated / distorted characters?
I see it makes some string values for each character like the "use hash values", but those don't look like hashes. Will it work with images having characters at different positions as there there's no box region like in "use hash values", so i assume it auto determines position of characters or sets it as fixed from the sample images?..
In some of your definitions i see you make the images black & white (thresholded) first with the filters, for others not.. so does this method rely on having the characters in exactly same colors like the "use hash values" relies on to make same hash value?
What the "must match atleast by X%" and "Add if match is less than X%" options mean and do?
I will appreciate your input as that will make me understand this solving method and help me make better solving definitions
Comments
The mask method works like this: It has to have a segmentation filter (segmet, segment-rotate, segmentV) as last filter to work. As long as you can get a filter set done which can seperate the chars in the image, then this is a good method for you. It is not working for recaptcha e.g. as those captchas have always intersecting chars in it who are hard to segment into single chars.
Anyway this mask is nothing more than a fingerprint of each char and with a lot samples it is a good way to let it solve fast.