Multiscript Handwritten Numeral Recognition

Hidden layer plays an important role in the performance of Back Propagation Neural Network especially in the case where the database is large and accuracy is the chief criteria. A lot of work has already been done with a single hidden layer and multiple hidden layer and it has been found that increasing the hidden layer also increases the accuracy performance slightly. But the drawback of increasing the hidden layer is that it also increases the network complexity as well as time. This project analyses the importance of proportionality of neurons in Multiple Hidden Layers in Back Propagation Neural Network. Usually in any common neural network based problem, maximum two hidden layers are enough to train the network properly and get the desired output. But in very complex problems where training data has got similarity among the data set then accuracy are the chief criteria. Specially in case of Multiscript Numeral Recognition where the shapes of different numerals resemble different values when analyzed. In this Project, a five hidden layer approach has been made in order to get the most possible accuracy in the Multiscript Pin Code Recognition system using PYTHON. For this application the proportionality of hidden layers are semi optimized to precise level to get the best results.