Estimating Cyclone Intensity on INSAT 3D Infrared Imagery Using Deep Generative Models

Determining cyclone intensity is crucial for disaster preparedness and the allocation of resources, making it an important aspect to focus on. This approach presents a hybrid model that leverages Variational Autoencoders (VAEs) to extract features and employs Generative Adversarial Networks (GANs) to enrich the available data through the integration of deep learning techniques. By tackling data shortages, the model can concentrate on essential latent features while enhancing interpretability. The proposed model aims to replace reliance on manual methods, such as the Dvorak technique, while still achieving high prediction accuracy. The potential of generative models in meteorological contexts will be showcased using the INSAT 3D datasets for experimental validation. To enhance the precision of cyclone intensity estimation, this article introduces a hybrid deep learning model that combines VAEs and GANs. This study enhances the advancement of generative models aimed at more accurately estimating cyclone intensity and also establishes a foundation for future exploration in meteorology and disaster management.