Tiara Putri Sakira, Muhammad Fajar, Andi Baso Kaswar, Satria Gunawan
Cassava plays an important role as a staple food source in Indonesia because of its high production level. However, the productivity of cassava plants can be negatively impacted by pest and disease attacks on the leaves. Although the symptoms of cassava leaf disease can often be identified visually, more expertise is needed to differentiate one disease from another. To overcome this challenge, Convolutional Neural Networks (CNN) was used for disease classification in cassava plants, using a dataset of 10,423 images with data augmentation. The results show that the most effective model is the EfficientNetB0 model, which utilizes a learning rate of 0.0001, 25 epochs, and a dropout rate of 0.4. This model successfully classified cassava plant diseases with an impressive accuracy of 98.09% during the validation stage and achieved a testing accuracy of 90.74%. This research shows the potential of CNN and transfer learning techniques in classifying cassava plant diseases accurately. © 2024 Author(s).
Departement of Computer Engineering, State University of Makassar, Makassar, 90224, Indonesia
