Natural disasters act as a serious threat globally, requiring effective and efficient disaster management and recovery. This paper focuses on classifying natural disaster images using Convolutional Neural Networks (CNNs). Multiple CNN architectures were built and trained on a dataset containing images of earthquakes, floods, wildfires, and volcanoes. A stacked CNN ensemble approach proved to be the most effective, achieving 95% accuracy and an F1 score going up to 0.96 for individual classes. Tuning hyperparameters of individual models for optimization was critical to maximize the models' performance. The stacking of CNNs with XGBoost acting as the meta-model utilizes the strengths of the CNN and ResNet models to improve the overall accuracy of the classification. Results obtained from the models illustrated the potency of CNN-based models for automated disaster image classification. This lays the foundation for expanding these techniques to build robust systems for disaster response, damage assessment, and recovery management.

Key Contributions

• Comprehensive Evaluation of CNN Architectures
• Stacked Ensemble Learning Approach
• Hyperparameter Tuning for Optimization
• High Accuracy and F1 Score
• Potential for Real-World Disaster Management Applications

Conclusion

By performing this research work we have demonstrated a novel approach for classifying disaster images using Convolutional Neural Networks (CNNs). The results generated by our CNN-based models including Resnet and stacked CNN ensembles, have achieved an accuracy of over 95% in classifying key disasters namely earthquakes, floods, volcanoes and wildfires. As per our analysis, the stacked CNN ensemble model was built using the basic CNN and a Resnet Architecture as base models and finally Ensembling using XGBoost Classifier has emerged as the top performer. We can state that combining and leveraging the strengths of individual base models to create a new hypermodel helps in producing better results. Other than that optimal model tuning and selecting the best hyperparameters like learning rate, number of epochs and batch size were also important to maximize accuracy. Our research is well-established on the concepts of deep learning and CNNs for disaster image classification. The techniques we propose can be further utilized in the development of automated systems for disaster response, damage assessment as well and recovery management. Additionally a scope of disaster classification can be road segmentation. This involves providing survivors with a precise road or exit path after a catastrophe has occurred, thereby improving post-disaster recovery efforts. Despite the usage of large disaster imagery dataset, there are certain limitations. The dataset may not represent the full diversity of disasters in real life. Additional tuning might further improve the model performance. Future research should concentrate on increasing the size and diversity of the dataset, combining disaster image segmentation with classification, and deploying these models in real-world disaster management systems. Model robustness and generalization testing will also be essential. More robust disaster management systems can be developed to mitigate risks and improve preparedness globally by addressing these limitations and building on current CNN advances. This study lays the groundwork for progress towards that goal.