Improving the YOLO v4 algorithm applied to the powdery mildew and downy mildew monitor system on cucumber plants

Abstract

This paper presents the research and design of the powdery mildew and downy mildew monitor system on cucumber plants. The system is designed based on Raspberry PI embedded computer and 5MB PI Camera. We have improved the YOLO v4 algorithm, then integrated it into this system to identify powdery mildew and downy mildew on cucumber plants. The test results of the improved YOLOv4 algorithm with the sample data set gave an accuracy of 83.43%, which is 0.37% higher than the original YOLO v4-tiny algorithm. The monitoring system initially operates stably with Raspberry PI hardware in the greenhouse at Vietnam National University of Agriculture.