Abstract

Rain Guard is an intelligent weather prediction system designed to detect rain conditions using deep learning and computer vision techniques. The system processes sky images through a Vision Transformer (ViT) model that classifies weather into categories like sunny, cloudy, and rainy with high accuracy. Input images are resized, normalized, and augmented to improve generalization during model training. The classified result is further enhanced using a generative AI model from Google Gemini, which provides contextual weather descriptions. Upon detecting adverse conditions, Twilio’s SMS API sends real-time alerts to users. The entire application is built with a Flask backend and an HTML/CSS frontend, ensuring ease of access and deployment. This lightweight and scalable solution offers an efficient alternative to traditional forecasting systems, especially for areas with limited meteorological infrastructure. It minimizes dependency on expensive hardware and complex simulations. The modular design supports integration into IoT and mobile-based environments.