Detection and Identification of Abaca Plant Pests and Diseases using Computer Vision and Deep Neural Network

Abstract

This study developed a prototype for detection and identification of abaca plant pests and diseases. The creation of the prototype comprised both hardware and software development. In software development, the process was divided into two, namely model development and user interface (UI) development. During the model development, the pre-trained model of InceptionV3 CNN architecture was utilized to train the two classes of abaca diseases such as the abaca mosaic and abaca bunchy top virus (ABTV) and three classes of pests particularly the Brown Aphids, Slug Caterpillar, and Corm Weevil. The final model was integrated in the developed UI. The UI was developed using three different frameworks such as OpenCV, TensorFlow Lite and Flask. Moreover, for hardware component, the Raspberry Pi 4 was utilized as major component and as a microprocessor of the whole system. Upon testing the prototype, the performance achieved an accuracy rate of 94.6%. Its commendable performance shows its possibility to be deployed in a real-world detection and identification of abaca plants pest and diseases.