Peach Leaf Diseases Identification Using Convolutional Neural Network and Fastai Framework

Abstract Peach fruit is an ephemeral tree native to the locality of Northwest China between Tarim Basin and Kunlun Mountains' north, where it was first cultivated. Peaches have quite several health benefits such as improvement in digestion, smoother skin, and relief from allergies. However, the growing quality of peaches is a challenging task due to the common peach diseases such as brown rot, bacterial canker, peach scab, and many more. Traditionally, farmers identify peach diseases by naked eyes from the peach leaves, which is often inaccurate and unstable due to the lack of professional knowledge. Therefore, this research developed an integrated algorithm using a convolutional neural network (CNN) and Fastai Framework, known as CNN-F for peach diseases identification. To experiment with the developed algorithm, 2,657 images of diseased and non-diseased (healthy) peaches leaves are collected from the PlantVillage dataset. Using the integrated CNN-F with different models like ResNet-34, AlexNet, VGG16, and ResNet-50, the result accuracy rate has reached 94.12%, 91.12%, 92.14%, 93.23% respectively. Therefore, the application of CNN-F with ResNet-34 provides the best performance to identify the diseased leaves of peaches. This indicates that the developed algorithm can improvise the efficiency of peaches production by incorporating this algorithm to assist the farmers to cultivate healthy peaches.