Building Machine Learning Models for Classification of Text and Non-text Elements in Natural Scene Images

Abstract

Computer vision aims to build autonomous systems that can perform some of the human visual system’s tasks (and even surpass it in many cases)among the several applications of Computer Vision, extracting the information from the natural scene images is famous and influential. The information gained from an image can vary from identification, space measurements for navigation, or augmented reality applications. These scene images contain relevant text elements as well as many non-text elements. Prior to extracting meaningful information from the text, the foremost task is to classify the text & non-text elements correctly in the given images. The present paper aims to build machine learning models for accurately organizing the text and non-text elements in the benchmark dataset ICDAR 2013. The result is obtained in terms of the confusion matrix to determine the overall accuracy of the different machine learning models.