Empirical study on two-class image classification

Abstract

Aim of this research is to study the two-class classifiers. A series of experiments were carried out using the three basic colour models - RGB (red, green, blue), HSV (hue, saturation, value), and the image temperature based model CCT (Correlated color Temperature); and feed forward artificial neural network model, CNN (Convolutional neural network). Combinations of these models were also tested to check which combination or the individuals prove to be relevant for indoor and outdoor scene classification, which has two distinct classes. Among colour models based experiments, the model with the combination of average RGB and average HSV achieved the highest accuracy of 80.95% and the model with the combination of average RGB and average CCT achieved the lowest accuracy of 52.38%. CNN with Adam and Stochastic gradient descent optimizers showed the best performance among the considered eight optimizers.