A Bag of feature approach to visual recognition and image classification

Abstract

This paper presents the improvements of bag-of-features based visual recognition and also looking into visual classification of objects and will develop a system that, given a number of images, can classify them into several categories based on shared visual feature. We start with a large-scale evaluation of a bag-of-features image and video classification framework. Features most useful for visual recognition in real-world conditions are chosen carefully and different order less or locally order less feature distributions are evaluated. We will outline the key aspects of research, covering the background of computer vision, specifically image classification, up to the current research. We will implement an existing method outlined in a suitable research paper which we will then test against that of the current art. The classifier will also be tested with images that are not photographs such as drawings, paintings and other forms of artwork. These images will include objects that are known to be identifiable by the classifier hence we can test the capability precisely for this kind of i/p. This research plans to im-prove on the success rate of organizing images by testing novel types & tweaking existing ones.