<![CDATA[This study aims to implement Deep Learning methods for early detection of extreme weather disasters based on satellite cloud image analysis. The dataset consists of multi-spectral imagery obtained from the Himawari-8 satellite, covering various atmospheric conditions. The proposed approach employs two main models: Convolutional Neural Network as the baseline model and Vision Transformer as the comparative model. The research methodology includes data preprocessing, model training, evaluation using accuracy, precision, recall, and F1-score metrics, and model interpretation using Explainable AI techniques. The results indicate that the Vision Transformer outperforms the CNN model, achieving an accuracy of over 92%. Furthermore, Grad-CAM visualization demonstrates that the model effectively identifies cloud regions associated with extreme weather phenomena. This study contributes to the development of an accurate and interpretable cloud-based early warning system, with potential applications in disaster mitigation, particularly in regions prone to extreme weather such as Indonesia.]]>
Copyrights © 2024
