Abstract

With an upsurge in the rate of data production, pervasive usage of cameras for automation and surveillance and the requirement of visual input for artificially intelligent devices all across the globe, there has been a rapid increase in the mass of image data being generated today. This gives rise to the essentiality of automated image processing required to simplify image related tasks. Automated image processing bridges the gap between the human visual system and the pixel level data of images. Deep Convolution Neural Networks are being deployed expansively to analyze, detect and classify images for a diverse number of tasks. These neural networks, similar to the human neural network, contain neurons with learnable weights and biases, which are trained to identify and classify different objects or features across the image. This paper presents a functional implementation of image recognition using a small convolutional neural network, proposing less complexity and yielding good classification accuracy for all tested data sets.