Published Paper Details:

Water body identification and monitoring are crucial in Earth observation and environmental monitoring efforts. These bodies, including natural lakes, rivers, reservoirs, and human-made ponds, hold significant ecological and practical value, impacting ecological balance, resource management, and disaster risk reduction. While methods like NDWI and MNDWI have been traditional tools for delineating water bodies from satellite data, the rise of machine learning offers new avenues for accuracy enhancement. This project focuses on using machine learning for water body extraction from remote sensing data, aiming to improve detection accuracy beyond traditional methods. Through model evaluation, the study seeks to advance current practices, benefiting resource management, environmental conservation, and disaster readiness. The trained model shows steady improvement over 60 epochs, starting with an accuracy of 63.39% and a loss of 0.5331 in the first epoch, eventually reaching approximately 81.14% accuracy and a loss of 0.1505 by the final epoch. This progression indicates the model's ability to learn and adapt, capturing underlying data patterns effectively. Validation accuracy also increases consistently, reflecting the model's ability to generalize well. The convergence of training and validation metrics signifies the model's stability and reliability in making predictions on unseen data.