A Real Time Intelligent Image Based Document Classification Using CNN and SVM

Abstract

Artificial Neural Network with multilayers is referred to as Deep Learning. As it is able to handle large amount of data, deep learning has become very popular in recent years. Classification, object tracking, estimation of different poses, text detection & recognition are various fields where deep learning is applied. Convolutional Neural Network(CNN) being one of the most important deep learning methods, has surpassed the traditional methods of classification.This paper presents a Supervised CNN, which is a machine learning algorithm used to classify images into different categories. In a few online registration processes, documents like Aadhaar,PAN,mark sheets etc need to be uploaded. Manual mistakes during upload can be committed as in; in place of Aadhar,some irrelevant document can be uploaded. Manual intervention to detect such mistakes is tedious.Hence we present an Intelligent Image based Document Classification using Machine Learning, where we have applied CNN to predict if the uploaded document is valid or invalid. The system is implemented on 900 images among which 500 are Aadhar images and remaining are PAN. The generated model recorded 90% accuracy to predict the Aadhar document and 93% in case of PAN.The same dataset is compared with Support Vector Machine(SVM) which displayed 88.88% overall accuracy.