A Framework for the Classification Task of Recognizing Weather Condition in an Image using Supervised Learning Methods

Abstract

The supervised learning methods are widely used in research area to predict very useful things and inference something from data. In this paper, we aim to predict weather condition in a given image/picture by using advanced supervised learning algorithms and various descriptors. The prediction of weather conditions from the image can be challenging and complicated in situations where large data sets are considered. In addition, for our purpose, we have separated train, validation and test images. And, in other words, we will classify an image into five classes as Cloudy, sunny, foggy, wet and snowy. The proposed methodology consists of four steps, pre-processing of the image was done in the first phase, extraction of the different features in the second phase. n the third phase of our methodology classification was carried out by applying the specified classification models to an input image. Finally, validation was performed for given classification results. The ultimate purpose of the knowledge obtained from the study is to developing a framework for the classification of recognizing weather condition using supervised learning methods are CNN, SVM, Random Forest, and Decision Tree.