Interpolating destin features for image classification

Abstract

This paper presents a novel approach for image classification, by integrating advanced machine learning techniques and the concept of feature interpolation. In particular, a recently introduced learning architecture, the Deep Spatio-Temporal Inference Network (DeSTIN) [1], is employed to perform feature extraction for support vector machine (SVM) based image classification. The system is supported by use of a simple interpolation mechanism, which allows the improvement of the original low-dimensionality of feature sets to a significantly higher dimensionality with minimal computation. This in turn, improves the performance of SVM classifiers while reducing the computation otherwise required to generate directly measured features. The work is tested against the popular MNIST dataset of handwritten digits [2]. Experimental results indicate that the proposed approach is highly promising, with the integrated system generally outperforming that which makes use of pure DeSTIN as the feature extraction preprocessor to SVM classifiers.