A Review of Agriculture Crop Diseases Detection Using Deep Learning

Abstract

Crop diseases has been causing a lot of loss in agriculture sector. The fast and accurate diagnosis of crop diseases is crucial in preventing and limiting loss from the crop diseases. To achieve this goal, method such as deep learning can be used to detect crop diseases. In this study, we review and study the performance of three convolutional neural network model, which is VGG16, VGG19 and Resnet50 model to classify crop diseases. Transfer learning with full connected layer are used, to shorten and decrease the training time and images needed. The dataset used for the experiments is from online plant disease database which is Plant Village Dataset. 210 images of tomato leaves are used in this research. The precision, recall, accuracy and F1-score are calculated for performance evaluation. The result show that Resnet50 perform the best compared to the other deep learning models with accuracy of 92%.