Optimized YOLOv8 Model for Accurate Detection and Quantification of Mango Flowers

Abstract

Mangoes are widely cultivated and hold significant economic value worldwide. However, challenges in mango cultivation, such as inconsistent flowering patterns and manual yield estimation, hinder opti mal agricultural productivity. This study addresses these issues by leveraging the You Only Look Once (YOLO) version 8 object detection technique to automatically recognize and quantify mango flowers using image processing. This research aims to develop an automated method for detecting and esti mating mango yields based on flower density, representing the early stage of the plant growth cycle. The methodology involves utilizing YOLOv8 object detection and image processing techniques. A dataset of mango tree images was collected and used to train a CNN-based YOLOv8 model, incorpo rating image augmentation and transfer learning to improve detection accuracy under varying lighting and environmental conditions. The results demonstrate the model’s effectiveness, achieving an av erage mAP score of 0.853, significantly improving accuracy and efficiency compared to traditional detection methods. The findings suggest that automating mango flower detection can enhance preci sion agriculture practices by reducing reliance on manual labor, improving yield prediction accuracy, and streamlining monitoring techniques. In conclusion, this study contributes to the advancement of precision agriculture through innovative approaches to flower detection and yield estimation at early growth stages. Future research directions include integrating multispectral imaging and drone-based monitoring systems to optimize model performance further and expand its applications in digital agri cu