Oil Palm Leaf Disease Detection on Natural Background Using Convolutional Neural Networks
Anindita Septiarini, Hamdani Hamdani, Eko Junirianto, Mohammad Sofyan S. Thayf, Gandung Triyono, Henderi Henderi
Abstract
Oil palm plant diseases typically manifest themselves on the leaves, resulting in reduced crop quality. It is necessary to solve this issue as the need for premium-quality palm oil keeps growing. Despite the fact that various automatic detection models for oil palm leaf disease have been developed, their performance was frequently inadequate due to the similarity of class characteristics. This work proposes a method that automatically detects the oil palm leaf disease on a natural background to distinguish between infected and healthy leaf classes. The method was developed using deep learning based on Convolution Neural Network (CNN) model. The private dataset consists of 600 oil palm leaf images (300 healthy and 300 infected) on a natural background. In order to decrease the computation time, pre-processing was carried out, which consists of resizing and normalizing the image, followed by augmentation. Augmentation was applied by rotation, flip, shear, and zooming techniques. Furthermore, the CNN model was employed to detect oil palm leaf disease using Tensorflow 2.5.0 framework with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$224\ \times\ 224$</tex> input data. The proposed method successfully achieved the highest performance, revealed by the accuracy value of 1.