Dataset for flood area recognition with semantic segmentation

Floods are natural disasters that repeatedly occur in Indonesia, causing substantial material losses and claiming many lives. Meanwhile, social media data has emerged as a valuable resource for analyzing user behaviour and interests, and its use for flood-related information is increasing. In this paper, we present a flood dataset collected from Instagram Reels, which consists of videos depicting flood events in Parepare. Every video was collected from different areas, time conditions and viewpoint, and converted into image form. The data set includes 7248 images. Images undergo preprocessing to ensure a clear depiction and differentiation of the flood event from the surrounding elements. Annotations given to each object, using a different color label, facilitate recognition and understanding of various computer vision applications. Overall, this flood dataset is a valuable resource for computer vision research, especially semantic segmentation method and promotes the development of algorithms for flood area identification and object recognition in flood-affected areas.