DESIGN AND IMPLEMENTATION OF DEEP LEARNING METHOD FOR IMAGE MINING FOR FACE DATABASE

Abstract

We present a project that proposes an algorithm which is able to segregate random pictures into organized groups. We use the technique of unsupervised learning i.e. clustering for doing so. Our effort makes it possible to cluster all images with respect to the categories we create. Here we will focus on one ofthe categories i.e. poses. We are using five datasets named Georgia Tech face database, Yale Face database, CAS-PEAL, MIT-CBCL and Kohn Kanade. However, in this particular piece ofwork, we will use Georgia Tech face database. We are focussing on mining the unlabelled images having various poses into separate groups. For this we are applying the unsupervised learning technique ofmachine learning. We are making clusters that will gather similar poses ofthe subjects. Hierarchical Clustering method is being used for this purpose. The images are inserted into the algorithm; they are then pre-processed so that they all align over a specific set of resolution and image type. After preprocessing the algorithm.form clusters and sends images to their respective clusters.