Identification of Durian Leaf Disease Using Convolutional Neural Network
Jay Al Gallenero, Jocelyn F. Villaverde
Abstract
Durian leaf disease is a severe agricultural issue in many parts of Southeast Asia, especially in the Philippines. The conventional way of detecting these diseases was through manual human eye monitoring and laboratory testing, which are crucial and time-consuming in controlling the development of these diseases. When it is present, the durian tree will not produce crops. Hence, this study has shown that a portable device embedded with the Duri Premium application was able to identify the durian leaf disease and provide treatment using Convolutional Neural Network MobileNet, a pre-trained model which suffices visual processing and would significantly help reduce the economic losses associated with the diseases such as algal spot, leaf blight, leaf spot, healthy leaf, and unknown. A total of seventy-five (75) samples were analyzed and used a confusion matrix to calculate the system's accuracy, which is 93.333%. As a result, the technique efficiently identifies the diseases mentioned above on durian leaves.