ZennoLab

Automate everything

User Tools

Site Tools


Sidebar

Translations of this page:

en:addons:capmonster:learning

Module Creation Studio

This application allows to create your own captcha recognition modules for CapMonster 2 software.

New version is based on modern technologies and considerable differs from the previous version. The process of module creation may seem quite complicated at first sight, but we will describe all the details in help articles and video guides. If you still have questions after reading help files, post and discuss them on our forum, we will be glad to resolve your issues and add necessary information to our help taking into consideration your feedback.

Download latest x64 version

Creating your own recognition module

Creation of your own recognition module is divided into several stages. Here we offer brief overview of every stage with the link to full description. If something is not clear, ask on our forum in special section devoted to CapMonster, otherwise you may just loose your time.

Creating project

This stage is common - open the program, create new project, save it under the name you like and start working with the program. At first you should load sample images of captcha, for which you would like to create recognition module, and recognize them via services with 100% accuracy.

Step 1: Creating project

Collecting symbols

Captcha solving is actually recognition of symbols on captcha image, i.e. you should train your module to recognize captcha symbols. For it you need to collect all captcha symbols. Unfortunately, you have to do it manually. But in new version it is much easier - you should just click on captcha image marking center of a symbol.

Step 2: Collecting symbols.

Applying filters

You can significantly increase captcha recognition rate applying graphic filters to captcha images. For example, you can adjust contrast and distinguish certain color. The most useful filter is Resize, it allows to enlarge the image with small symbols and narrow image with large symbols.

Step 3: Applying filder

Setting mass centers

As we said above, captcha solving means recognition of captcha symbols, but before recognizing you should find them on captcha image. CapMonster is searching for symbols at special points - potentional mass cetners. At this stage you should also adjust frame size - the rectangle which fits the largest symbol in your captcha.

Step 4: Setting mass centers

Training module

When all required resources are loaded and all useful settings are configured you can begin training your module. For the first you should set core complicity depending on symbols image distortion. Complex core provides high recognition rate for a long time and simple core allows to process captcha quickly, but at low recognition rate. Then you should configure training settings. It is simple - than faster training, than lower recognition rate of your resulted module.

Training is the longest part of module creation process and in some cases may last almost a day. Fortunately, your presence is not required.

Step 5: Training module

Testing module and improving its quality

After training your module you can test how it recognizes captchas. At this step you will be able adjust some settings to improve recognition capabilities of already created module. You can also find out the reasons of some recognition errors and try to correct them.

Step 6: Testing module

Export of workable module

Module Creation Studio application only creates modules and stores them with captcha images and other non-useful information. So, after testing you should save created module in proper format to load it into CapMonster 2 application which recognizes captchas.

When module creation is finished and you are satisfied with test results, you should press “Save” and select the path on drive to save your module as file. After that you will be able to load and use it in CapMonster 2 for recognizing captchas or sell to someone.

Step 7: Exporting module

en/addons/capmonster/learning.txt · Last modified: 2021/01/23 21:14 by sergodjan