Optimizing Weather Pattern Recognition with MobileNetV3 Architecture

Weather image recognition has become an essential component in meteorology and environmental monitoring, enabling automation in weather pattern classification and improving decision-making processes. This study explores the use of the MobileNet V3 model, a lightweight and efficient convolutional neural network, for the classification of weather images into categories such as sunny, cloudy, rainy, and snowy. Leveraging transfer learning and advanced data augmentation techniques, the model effectively extracts robust features while maintaining computational efficiency. The performance of the model was evaluated over 12 epochs, achieving a validation accuracy of 91.88% and a validation loss of 0.4424, showcasing its ability to generalize well on diverse datasets without overfitting. The results indicate a consistent reduction in training and validation losses, highlighting the effectiveness of MobileNet V3's architecture for resource-constrained environments. Furthermore, the model's adaptability makes it suitable for real-time applications such as autonomous systems, disaster management, and agricultural automation. This study provides a comprehensive evaluation of MobileNet V3 for weather image recognition, emphasizing its efficiency, scalability, and accuracy. Future directions include exploring advanced optimization techniques and integration with other state-of-the-art architectures to enhance performance further. This research contributes significantly to advancing deep learning-based weather recognition systems for practical and impactful applications.