What is the most appropriate size for Training set Images for Handwriting Recognition

Question

I'm developing an offline handwriting recognition application using neural networks(Feed forward back propagation method). I'm confused about the normalized size of the training set images of the isolated characters. Currently i'm thinking of 32*32 pixels. Can please help me with this and are there any specific considerations to this matter. thank you!!

Answer 1

One of the more popular datasets is the MNIST Handwritten Digits set, which contains handwritten digits. This is a version for Matlab, and Sci-kit has a python machine-learning demo with it.

You will find loads of scientific papers using it as a benchmark.