Classifying Natural Disaster Scenes

Overview

The main objective of our is to minimize the horrible effects of that natural disasters have around the world. By utilizing the immense quantity of images contained in the LADI dataset in order to test different classifying models, our team will be able to come up with the best model, which we called the "Natural Disaster Classifier." We will use different sections of the LADI dataset as the inputted images for previously existing models, and this will allow for continued learning while we train multiple models using different variables like image pre-processing techniques, CNN architecture, and techniques to reduce overfitting. We will then keep track of the performance of each model being tested as well its specifications, parameters, and architecture, etc. Thanks to this, we will have accessible information about previously tested models and that will make it much simpler when deciding on the right model, which we will then test using the entire LADI dataset. This will result in a model that can accurately classify natural disasters, such as wildfires, hurricanes, floods, and earthquakes.

Community Benefit

In today’s world, natural disasters are occurring with an increasing frequency and intensity, which poses significant challenges to many communities around the world. The lack of accurate data worsens preparedness efforts and hurts swift response measures. Our classifier could help scientists to warn people with enough time to prepare for these events, enabling them to make informed decisions and implement proactive measures to safeguard their lives and their properties.

Team Members

• Jehad Ismail
• Juan Hernandez
• Alfonso (Fonsi) Avila
• Eric Smith
• Jacob Torres

Sponsors

Florida Atlantic University & Multimedia Lab