Detection and Identification of Abaca Diseases using a Convolutional Neural Network CNN

Abstract

To have an alternative way to detect and identify abaca plant diseases, the manual approach has been devised using the Raspberry Pi 4, Raspberry Pi H.Q. Camera, and Raspberry Pi LCD Monitor. The study used the Convolutional Neural Network-VGGNet-16 architecture. The model was trained using 300 training datasets wherein, for each class, there are 100 training samples. The researcher split the training data into 80% training and 20% validation data. The model achieved an accuracy rate of 88.9% and 91.7% of the average precision rate upon its testing. The implementation of this study can be advantageous for the early detection and identification of abaca plant diseases. The researcher also notes that the device can be applied for flying objects for a broader detection range and identification for future works.