In this research, we developed a method to detect leaf rust symptoms using transfer learning. Leaf rust is a plant disease caused by fungi from the Pucciniales group (formerly called Uredinales). Rust disease can impair plants’ physical performance and reduce the quality of melon growth. It can also deplete starch reserves in the roots and branches. Eventually, the disease may kill the trees. We obtained the dataset by collecting leaf images from the Institut Pertanian Stiper (INSTIPER) greenhouse in Yogyakarta. We collected images of 81 healthy leaves and 82 leaves affected by rust symptoms. Each image was divided into nine parts using a 3×3 grid to increase the data size. After splitting, the number of healthy leaf images was 729, and those with rust symptoms were 738. We used K-fold cross-validation to split the dataset into 5 folds. Four parts were used for training and one for testing. Using K-fold provides a comprehensive performance metric. For the model, we used transfer learning models such as Xceptiion, ResNet50V2, and EfficientNetV2S performed best. It achieved a Precision of 0.959, a Recall of 0.96, an F1-score of 0.9595, and Accuracy of 0.958.<\/p>\n
Authors:
\nAndrea Stevens Karnyoto, Claudia Rachel Wijaya, Harsawardana, Wiwin Dyah Ully Parwati, Muchammad Ali Nur Secha, Taufik Yusuf, Yasin Addafi, Najia Al Umri, dan Bens Pardamean.<\/p>\n