A Novel Approach for Image Classification using Annotations

Abstract

The syntactical exploitation of the data and searching of the images on the Web demands the inclusion of the supplement of some finite structures into the list of standards to enhance the accuracy of the search results. We propose a novel image classification technique, where images tag is used as only means to provide the pertinent information about the image. This tag is used in conjunction with the current Web standards to classify the uploaded images into a finite number of related groups. The categorization process is in two stages: preprocessing stage where data is extracted from the annotation tag of the image and training and classification stage where extracted data is trained into categories and new data is classified based on the trained data set. We have used Random Forest based classifier and IBk for experimentation and result evaluation. Our work relies on simple and extensible keyword-based query language and enables efficient categorization of images.