Klasifikasi Penyakit pada Citra Buah Jeruk Menggunakan Convolutional Neural Networks (CNN) dengan Arsitektur Alexnet

Abstract


 
 
 Citrus fruit is a plant that is very susceptible to disease. Diseases that often attack citrus fruits are usually in the form of spots on the fruit. Diagnostics of citrus fruit diseases are usually carried out by experts manually which can cause the results to be subjective. Not all farmers are experts in diagnosing citrus fruit diseases. Therefore, this study proposes a system for diagnosing citrus fruit diseases using computer vision based on deep learning. So that the model can be used on computers with limited resources, this study proposes the Alexnet model, which is relatively light but has proven excellent accuracy in classifying several datasets. The dataset used is citrus fruit disease images of 1790 images which are divided into 4 classes, namely blackspot, canker, grenning, and healthy fruit. The best results achieved with a scenario of 90% training data and 10% validation data are with an accuracy of 94,34%, a precision of 93,0%, a recall of 94,0%, and an F1-score of 95,0%. The best results are obtained with a combination of dropout, batch normalization, and fully-connected layer scenarios in the classifier layers section.