Flood Detection Using Convolutional Neural Networks and Satellite Imagery

Flooding is a devastating natural disaster that causes significant damage to both the environment and human settlements. This work focuses on detecting the occurrence of flooding to enable timely intervention, thereby mitigating its consequences and enhancing resilience in flood-prone areas. The proposed system utilizes Convolutional Neural Networks (CNNs) to classify satellite images into two categories: ’Flooded’ and ’No Flood.’ A comprehensive dataset of satellite images has been curated for training, validation, and testing purposes. To im- prove model performance, various preprocessing techniques such as noise reduction and normalization are applied. Additionally, data augmentation methods including image flipping, rotation, and scaling are employed to enhance the diversity and quality of the training set, ensuring the model can effectively handle variations in input images. A userfriendly web application has been developed using Flask, enabling users to register, log in securely, and upload images for flood detection. The interface provides real-time predictions, classifying images as ’Flooded’ or ’No Flood’ while also displaying a confidence score to enhance trust in the system's outputs.