Classification of Paddy Leaf Disease Using MobileNet Model
Fauzan Masykur, Kusworo Adi, Oky Dwi Nurhayati
Abstract
Paddy is a rice-producing plant as a staple food raw material for most of the population in Indonesia. The rice planting process cannot be separated from pests and diseases that can cause crop failure and impact on rice production stocks. Currently, the process of monitoring the development of rice plants is done conventionally by direct observation by farmers. Meanwhile, in this paper, a classification model and determination of object detection in rice plants infected with pests and diseases is proposed based on leaf color. Leaf classification applies convolution conditional network algorithm with mobilenet as the architecture. The leaf image dataset was taken from public data from kaggle.com by involving 4 classes, namely healthy leaf images, brownspot, hispa and leafblast images. In each class the leaf images are grouped into 2 groups, namely 70% training and 30% validation. On the mobilenet backbone, it produces 97% accuracy and 22% loss with the time required for 195s 19s/step in 1 epoch round.