REAL-TIME IDENTIFICATION OF MAIZE PLANT AND WEEDS USING MACHINE LEARNING MODEL IMPLEMENTED VIA YOLO v5

Abstract

Agriculture is as old as the existence of humanity. Early humans engaged in fruit gathering and hunting for food. Agriculture has improved over the years and the need for effective agricultural practices has become a growing concern for farmers. Maize is a staple food in many countries and serves as a primary source of carbohydrates. It is versatile and used for human consumption, animal feed, and industrial applications. Weeds are one of the major obstacles faced in crop production. Weeds compete with plants for nutrients, space and sunlight, affecting the quality of crop yield. This paper aims to use machine learning to train a model to identify weeds and maize plants on the farmland. This research focuses on implementing a real-time detection system using the YOLOv5 (You Only Look Once version 5) deep learning model. The Machine learning model was developed using a training set of about 3000 images of maize plants obtained from the farmland. The obtained data were annotated via makesense.ai and the labelled dataset was divided into training and testing sets. An overall accuracy of over 90 per cent was achieved with the implemented model and a mAP value of 0.95, demonstrating the efficacy of our approach for identifying and differentiating a maize plant from the weeds.