Classification of Images in Bad Weather Conditions with Convolutional Neural Networks

Weather conditions are one of the major factors significantly influencing the daily lives of individuals. Unfavorable weather conditions adversely affect their lives and directly impede the progress of the subsequent image-processing steps necessary for real-world vision tasks such as object detection and autonomous driving. For this reason, the correct classification of the weather conditions is of great importance. Although traditional classification methods achieve high accuracy in various tasks, they cannot achieve the same success in classifying weather conditions. In this paper, we propose a novel convolutional neural network (CNN) framework for the classification of weather conditions with high accuracy. The proposed network outperforms the existing methods with 95.50% accuracy for a classification problem with five different scenarios.