OCR optical character recognition using neural networks


- To develop a software tool that automatically recognizes numeric characters (0-9) designed using the hand (mouse pointer) by the users.


- Software tool that provides a graphical user interface to design the numeric characters and to display the character recognized by the software tool. The software tool implements different architectures of neural networks (classifier and (associative memory + classifier)) with different activation functions profiles. The different architectures of the neural networks have been trained and tested with 50 (5x10) and 500 (5x100) samples of the numeric characteres.

- The MATLAB source code and documentation of the application is available in Github. A video demonstrating the software tool is presented next:

- Software tool developed in collaboration with Joao Duarte during the Adaptive Computing and Diffuse Systems course.