Classification of Grape Leaf Disease Using Convolutional Neural Network (CNN) with Pre-Trained Model VGG16
Yohan Rayhan, Djoko Budiyanto Setyohadi
Abstract
Grape is an agricultural commodity that has high value with price in Indonesia is Rp 15,000 – Rp 60,000/Kgs. Plant disease is something that can decrease the productivity of the grape plant. Black rot, black measles, leaf blight are diseases that are often found on grape leaves. This disease has visible symptoms so farmers can take precautions. The process of identifying leaf disease is very important to determine what steps the farmers can take. With Technology utilization, the identification process becomes faster and more accurate. The total dataset is 1,801 divided into 4 classes and 3 categories of data and used Convolution Neural Network (CNN) with pre-trained VGG model. The result is very good because from this model obtained accuracy 97.9% and loss 6.5% for training data and 93.3% accuracy and loss 13% for validation data in the tenth epoch. In Testing analysis, the analysis using confusion matrix with every class has 8 images obtained 100% accuracy of recognize for class black rot, healthy leaf and leaf blight but 87.5% accuracy in black measles class, because only recognize 7 images from 8 images.