Improving Image Classification on Collective Response by Using Unified Extreme Learning Machine Techniques

Abstract

The most crucial component of computer vision is Image classification. Image classification describes to a procedure in digital image analysis that can be used to categorize an image corresponding to its graphical content. Image Classification provides away to make visual content discoverable via search.  This paper puts forward the idea of improving the image classification by using the techniques of Extreme Learning Machines (ELM). ELM, are models which produce good generalization performance and also has additional features that help learn thousands of times faster than typical  and traditional networks which are trained using back propagation. BP based algorithms play dominant roles in training the feed forward neural networks (FNNs).The proposed system consists of two kinds of classifications: the binary ELM learning stage layer and the multiclass layered ELM (ML-ELM). The feature mapping of ML-ELM is done by repeating and alternating the feature map layer and maximum pooling operation.