Increased accuracy in predictive models for handwritten character recognitionwill open up new frontiers for optical character recognition. Major drawbacksof predictive machine learning models are headed by the elongated training timetaken by some models, and the requirement that training and test data be in thesame feature space and consist of the same distribution. In this study, theseobstacles are minimized by presenting a model for transferring knowledge fromone task to another. This model is presented for the recognition of handwrittennumerals in Indic languages. The model utilizes convolutional neural networkswith backpropagation for error reduction and dropout for data overfitting. Theoutput performance of the proposed neural network is shown to have closelymatched other state-of-the-art methods using only a fraction of time used bythe state-of-the-arts.
展开▼
