Utilization of DCT Coefficients for the Classification of Standard Datasets in Cloud/Edge Computing Environment

In this work we propose efficient deep neural networks for classification that are well suited to the edge computing and cloud computing environment. These environments inherently have to deal with bandwidth limitations and bounded computational resources. Our proposed methods tend to reduce the bandwidth requirements and reduce the computational costs for running these deep learning algorithms. We extensively utilized the Discrete Cosine Transforms (DCT) to exploit the redundancy in the image datasets. In this work we present a deep neural network that predicts the most significant DCT coefficients for an image and then employs these important DCT coefficients for classification purpose. This makes the deep neural network to achieve classification by processing much less input information. Broadly two approaches were used for classification purpose. In the first approach classification was done by employing the most significant DCT coefficients and in the second approach low resolution images, constructed from a limited number of DCT coefficients were utilized for classification purpose. The experiments were performed on well-known greyscale and RGB image datasets like FASHION MNIST, CIFAR-10 and CIFAR-100. VGG-16 architecture is mainly used for classification. The experiments showed promising results and the classification accuracies achieved were almost the same as that achieved by full resolution images.