A Diversity-optimized Deep Ensemble Approach for Accurate Plant Leaf Disease Detection

TL;DR

The paper addresses the challenge of reliably selecting high-quality deep ensembles for plant leaf disease detection, where traditional Q-diversity metrics often misalign with ensemble accuracy. It introduces the Synergistic Diversity (SQ) framework, which uses a focal-model-based negative-sample analysis to compute two components, SQ-ε and SQ-α, and combines them as $SQ = w_\epsilon \cdot SQ-\epsilon + w_\alpha \cdot SQ-\alpha$ with default weights $w_\epsilon = w_\alpha = 1$. Through extensive experiments on a Plant Leaf Dataset (~80,000 images), SQ-based ensembles consistently outperform traditional metrics, achieving up to 99.80% accuracy with compact ensembles and showing a strong positive correlation with accuracy ($0.549$). The results demonstrate that SQ enables scalable, high-accuracy plant disease detection by effectively identifying complementary ensemble members, reducing computational costs, and improving reliability in agricultural settings. These findings support broader adoption of SQ-based deep ensembles for image-based plant health monitoring and disease management.

Abstract

Plant diseases pose a significant threat to global agriculture, causing over $220 billion in annual economic losses and jeopardizing food security. The timely and accurate detection of these diseases from plant leaf images is critical to mitigating their adverse effects. Deep neural network Ensembles (Deep Ensembles) have emerged as a powerful approach to enhancing prediction accuracy by leveraging the strengths of diverse Deep Neural Networks (DNNs). However, selecting high-performing ensemble member models is challenging due to the inherent difficulty in measuring ensemble diversity. In this paper, we introduce the Synergistic Diversity (SQ) framework to enhance plant disease detection accuracy. First, we conduct a comprehensive analysis of the limitations of existing ensemble diversity metrics (denoted as Q metrics), which often fail to identify optimal ensemble teams. Second, we present the SQ metric, a novel measure that captures the synergy between ensemble members and consistently aligns with ensemble accuracy. Third, we validate our SQ approach through extensive experiments on a plant leaf image dataset, which demonstrates that our SQ metric substantially improves ensemble selection and enhances detection accuracy. Our findings pave the way for a more reliable and efficient image-based plant disease detection.

Figures (3)

• Figure 1: Overview of Our Deep Ensemble Approach
• Figure 2: Scatter Plots of Q Metric Scores and Ensemble Accuracy on The Plant Leaf Dataset: the legend on the right shows how different colors correspond to different team sizes.
• Figure 3: Correlation of Ensemble Diversity and Accuracy.