Aerial Image Quality Enhancement Using Enhanced Super-Resolution Generative Adversarial Networks for Rice Plant Disease Identification

Disease identification in rice plants is one of the essential elements in maintaining agricultural productivity in Indonesia. Tungro and blast are two diseases that often attack rice plants in Indonesia. Aerial imagery captured with the help of drones is a promising alternative for monitoring diseased areas on farms more efficiently. How-ever, aerial images produce low resolution when zoomed in, limiting the ability to identify diseases. The enhanced super-resolution generative adversarial networks (ESRGAN) method can increase the resolution of aerial images so that the classification of rice plant diseases with the convolutional neural networks (CNN) classification method performs better. The test results show that identifying rice plant diseases with aerial image data has an accuracy of 86.96% when the image data is enhanced in resolution with ESRGAN, and its accuracy is 79.97% without ESRGAN.