Trust-Aware Disaster Image Classification using Monte Carlo Dropout Integrated with Graph Convolutional Networks and Deep Feature Extraction

Disaster management necessitates efficient and accurate classification of disaster images for reliable and effective response. In this paper, we propose TDC-GCN (Trust Disaster Classification using GCN), a novel approach that integrates trust estimation through Monte Carlo Dropout (MCD) with a Graph Convolutional Network (GCN) and deep feature extraction techniques for disaster image classification. We utilized a diverse dataset from Kaggle, encompassing 6 distinct disaster categories, to train and evaluate our models. Our methodology incorporates a pre-trained ResNet50 model for feature extraction, followed by graph construction based on image similarities. The GCN model, comprising graph convolutional layers, is trained to classify disaster images into multiple categories while estimating trust using MCD for uncertainty estimation. Extensive experiments on a comprehensive disaster image dataset demonstrate the effectiveness of our approach, achieving significant improvements in classification accuracy compared to traditional methods. The innovative integration of GCN, deep feature extraction, and trust estimation techniques not only enhances accuracy but also provides a reliable and trustworthy framework for disaster management.