Conversion of handwritten forms to excel sheet

Abstract

In this project initially, we have done some literature survey and have done extensive secondary research to get an overall idea from scratch. In this process, we found out different ways to process the data. Such as OCR can be done by using classifier or by using neural networks. In classifier, there are a different kind of classifiers such as supervised: K nearest neighbor, minimum distance to mean, SVM, bays classifier, unsupervised: k means clustering. We studied neural networks also for optical character recognition as it is suitable for big data set and image recognition. We got in-depth knowledge about classifiers. On the other hand, we also had researched neural networks. There are different inbuilt software in different packages, which can also be used for optical character recognition. We studied thoroughly about all the tools and software which can be used for optical character recognition, which helped us to make the project success. For optical character recognition, we need to train a model which can identify the letters with high accuracy. Hence we collected the handwritten data. We collected these data from different people with different handwriting using a pen and paper. Only regular handwritten pages were not sufficient as people have cursive handwriting also, and it is difficult to identify those letters and word. Hence, we collected cursive handwriting data to made the model better. On the other hand, rather than collecting the data, lakhs of data can be generated by augmentation. Where you give a letter input and with distortion, it changes the orientation and curves shape of the letter and in this way numbers of data can be generated. For neural network, usually massive amount of data is required for processing