Comparative Evaluation of Inception V3 and YOLOv8 for Strawberry Plant Diseases Classification Using Deep Learning Models

Abstract

Plant diseases and pests threaten agricultural productivity, with leaf diseases causing major crop losses. Early detection is essential to mitigate these impacts. This study presents a system for detecting strawberry leaf diseases using deep learning-based Convolutional Neural Networks (CNNs) by utilizing two pre-trained models, Inception V3 and YOLOv8, to classify leaves as healthy or diseased. A custom dataset of 5,192 images, comprising one healthy class and four disease-infected categories (leaf blight, blotch, scorch, and spot), is used. Inception V3 achieved 93.8% accuracy, while YOLOv8 outperformed it with 95.4% accuracy, a mAP of 78.6%, and precision, recall, and F1-scores of 89%, 88%, and 89%, respectively. With a compact size of 12 MB and a rapid inference time of 10 ms per image, YOLOv8 is highly suitable for real-time applications. These findings highlight YOLOv8's potential to improve agricultural productivity and food security through precise and efficient disease detection.