Detection of Flood Images Posted on Online Social Media for Disaster Response

User-generated content on online social media (OSM) has several data mining applications, such as extracting useful information during disaster events. In this paper, we propose a CNN based model to detect flood images among several images posted on OSM. The proposed system consists of two levels of identification scheme. At the first level, a CNN model is used for detecting flood images, followed by a color based filtering to obtain the final detection. The proposed system is tested using real Twitter images collected during a recent flood event and the results are encouraging.