An efficient plant disease detection using transfer learning approach

2025-06-28

AI Summary

To address the challenges of early plant disease identification and reliance on labor-intensive manual monitoring, this paper proposes a lightweight object detection framework based on transfer learning. We conduct end-to-end fine-tuning of YOLOv7 and YOLOv8 on a multi-class leaf disease image dataset to achieve fine-grained localization and classification of bacterial, fungal, and viral diseases. A key contribution is the empirical validation of YOLOv8’s superior generalization capability in low-data agricultural scenarios. Model performance is rigorously evaluated using standard metrics: mean Average Precision (mAP), F1-score, Precision, and Recall. Experimental results demonstrate that the optimal model achieves an mAP of 91.05%, an F1-score of 89.40%, Precision of 91.22%, and Recall of 87.66%—substantially outperforming conventional approaches. The method exhibits high accuracy, computational efficiency, and strong scalability, offering a practical and deployable technical solution for intelligent plant protection systems.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Plant diseases pose significant challenges to farmers and the agricultural sector at large. However, early detection of plant diseases is crucial to mitigating their effects and preventing widespread damage, as outbreaks can severely impact the productivity and quality of crops. With advancements in technology, there are increasing opportunities for automating the monitoring and detection of disease outbreaks in plants. This study proposed a system designed to identify and monitor plant diseases using a transfer learning approach. Specifically, the study utilizes YOLOv7 and YOLOv8, two state-ofthe-art models in the field of object detection. By fine-tuning these models on a dataset of plant leaf images, the system is able to accurately detect the presence of Bacteria, Fungi and Viral diseases such as Powdery Mildew, Angular Leaf Spot, Early blight and Tomato mosaic virus. The model's performance was evaluated using several metrics, including mean Average Precision (mAP), F1-score, Precision, and Recall, yielding values of 91.05, 89.40, 91.22, and 87.66, respectively. The result demonstrates the superior effectiveness and efficiency of YOLOv8 compared to other object detection methods, highlighting its potential for use in modern agricultural practices. The approach provides a scalable, automated solution for early any plant disease detection, contributing to enhanced crop yield, reduced reliance on manual monitoring, and supporting sustainable agricultural practices.

Problem

• Detect plant diseases early using transfer learning
• Compare YOLOv7 and YOLOv8 for disease identification
• Automate monitoring to improve crop yield and sustainability

Innovation

• Transfer learning for plant disease detection
• YOLOv7 and YOLOv8 models fine-tuning
• High mAP and F1-score performance metrics