Handwritten Character Recognition

This NeuronDotNet sample demonstrates the use of backpropagation neural networks in pattern recognition applications.

• Handwritten Character Recognition
• Handwritten Character Recognition Source

This application consists a 'Controls' component which provides 'DrawingBox' and 'LetterBox' controls, and also defines a set of classes which can be used to develop applications which require handwritten (mouse-drawn) images as inputs. The 'OCR' component makes use of NeuronDotNet and the 'Controls' component to recognize mouse-drawn capital letters of English alphabet.

The application uses Backpropagation neural networks for recognition. For better accuracy, one neural network is used for distinguishing each pair of letters. Various instances of letters used for training are saved in 'Letters' directory and the trained networks are saved into 'Networks' directory.

The application provides user interface to allow users to add, remove and edit the instances in training set, and also allows users to build and train neural networks for customized training sets.

Screenshot - Recognition

Screenshot - Training

More Samples