Plant Disease Detection Based on YOLOv3 and YOLOv4
Apu Shill, Md Asifur Rahman
Abstract
Although the loss of yield from plant disease can be precluded with early detection and treatment, they cause significant economic losses every year due to the dearth of cost-effective expert knowledge. Not only they are a great threat to the agricultural economy but also they are pernicious to the balance of the natural ecosystem. Many studies have been conducted to develop expert systems to alleviate losses by detecting the diseases at their earlier stage. However, most of these studies are limited to one or two plant species and a few varieties of diseases. This paper introduces a computer vision approach using the latest state of art You Only Look Once (YOLO) algorithms in developing an expert system, which can accurately identify numerous plant diseases from a diverse set of plant species. The latest algorithms from the YOLO family namely YOLO V3 and YOLO V4 have been incorporated in this research and the experimental result presented in this paper clearly reflects on the accuracy and suitability of the plant disease detection system.