Klasifikasi Kondisi Langit menggunakan Algoritma Convolutional Neural Network dengan Arsitektur NASNet

Abstract

This research focuses on developing a sky image classification system using Convolutional Neural Network (CNN) with the NASNet architecture, renowned for its superior performance in image classification tasks. Sky images hold significant potential for identifying optimal locations for solar radiation absorption, with cloud coverage being one of the key influencing factors. In this study, the Singapore Whole Sky Imaging CATegories (SWIMCAT) dataset is utilized, comprising five categories: clear sky, pattern cloud, thick dark cloud, thick cloud, and veil. This dataset serves as the foundation for training and testing the NASNet model, enhanced with on-the-fly data augmentation techniques to boost its performance. Experimental results demonstrate that the proposed model achieves an outstanding accuracy of up to 99.37%. These findings highlight not only the reliability of NASNet architecture in image classification but also its potential for broader applications, including weather monitoring, solar energy management, and automated image processing. With its high accuracy and processing efficiency, this approach underscores its relevance in advancing AI-based technologies across diverse fields.