A Model for Improving Image Classification Using Convolutional Neural Network for Emergency Situation Reporting

Abstract: Artificial neural networks (ANNs) and Deep Learning have shown great improvement in recognizing and classifying objects and images. Convolutional Neural Networks (CNNs) can be exploited to automatically extract features from images using the hierarchical structure. Specifically, we investigate image classification in the context of every day small scale emergency events, where images associated with tweets during the target emergency events were used to train and test classification approaches. We present a method of classifying images using convolutional neural networks that perform classification in layers in order to determine whether or not the image relates to an emergency event and if the image is incident-related, then which category of the target emergency events it belongs to. Experiments on a home-grown dataset show that these methodologies can classify images into the different classes with an F1 score of 88.12%.

Keywords: Image classification, convolutional neural networks, social media, Twitter, Emergency events