This study presents a novel multimodal fusion technique for disaster identification in Bangla, combining text and image data using the "BanglaCalamityMMD" dataset. Employing DisasterTextNet, DisasterImageNet, and DisasterMultFusionNet, the approach addresses a key gap in Bangla disaster research.

Real-time system for collecting, processing, and distributing disaster-related media from social media platforms.

Enhanced disaster image classification using ConvNeXt with Squeeze-and-Excitation (SE) and Feature Pyramid Network (FPN). Built on the MEDIC dataset, this project aims to improve classification accuracy and address overfitting issues.