ocr4j is an open-source optical character recognition library fully implemented in Java programming language. Project is still in very early alpha stage and library is not yet very powerfull, it's recognition ability does not allow to use it serious applications.
Source codeCheck out latest source from subversion repository:
svn co https://svn.sourceforge.net/svnroot/ocr4j ocr4j
Demo appletThe following applet demonstrates the character recognition power of the library. The example uses a bipolar-sigmoid function based neural network with 3 neural layers. Feature extraction algorithm is a random line receptors. See documentation for more details. Network was trained to recognize latin capital block letters only.