Tomasz Gebarowski

Simple Digit Recognizer

A simple digit recognition application in Java using a concept of Neural Networks and back propagation algorithm.

The application is able to detect one character digits or symbols having some minor distortions.

My project is based on the algorithm described by Sacha Barber on the

Image Filter


It is important that before starting recognizing the digits, a neural network must be trained for a given set of data probes. Predefined probes can be loaded from the “Trainer” tab. It is possible to select from three different training sets (decimal, hexadecimal and binary numbers)

Image Filter

When a network is successfully trained, move to “Test (Drawing)” tab where you can draw a digit (possibly with some disortions) and test it against proper recognition. A set of input test probes can be extended in order to add some additional cases. It is important that after that the network must be retrained.

Image Filter