Student Research Reports

View Research Report

Weather greatly impacts daily life, influencing activities and global climate patterns. With climate change making weather more unpredictable, we face challenges such as storms, floods, water shortages, and landslides. These issues highlight the need for better weather monitoring systems. Clouds are essential in regulating temperature, driving the water cycle, and shaping weather. Since cloud types are directly linked to weather changes, identifying them helps predict short-term and long-term weather events, such as storms and extreme temperatures, improving preparedness. However, identifying cloud types requires expertise, and inconsistencies can affect the accuracy of weather predictions. This project aims to create a web-based application that uses machine learning to automatically classify cloud types. The goal is to make cloud identification more accessible, assisting in environmental monitoring and raising awareness of weather patterns. The project also promotes citizen science, encouraging public participation in improving weather forecasting and climate action. In this project, a machine learning model was trained to classify eight cloud types using a dataset of images. The model was tested with 175, 200, and 250 images to examine how dataset size affects accuracy. The best results came from using 175, 200, and 250 images, achieving a training accuracy of 96.5% and validation accuracy of 82.8%. The developed app, "Qmulo," performed well in classifying clouds, with Cumulonimbus clouds identified with 90% accuracy, Altocumulus at 88.57%, and Cirrus at 85.7%. When focusing only on correct classifications, Cumulus clouds were identified with 98.52% accuracy. While the application performed well overall, some cloud types, like Cirrocumulus and Stratocumulus, were classified with lower accuracy. This can be improved by expanding the dataset and selecting more representative images. The app shows potential in environmental monitoring and encourages public involvement in weather tracking. In conclusion, the "Qmulo" application demonstrates the effectiveness of machine learning in cloud classification, with room for future improvement in accuracy and broader public participation.