What question did this study set out to answer?

This research aims to enhance early detection of plant leaf diseases using advanced hybrid models for practical agricultural deployment.

May 12, 2026Open Access

Using Hybrid Pre-trained Convolutional Neural Networks and SVM Based VGG16, ResNet50, and DeseNet201 for Identifying Plant Leaf Disease

SYSura Yaareb RDRajaa Daami HMHasan Muayad

Key Points

This research aims to enhance early detection of plant leaf diseases using advanced hybrid models for practical agricultural deployment.
Developed a hybrid classification pipeline combining VGG16, ResNet50, and DenseNet201 with an SVM classifier.
Used HSV color space to improve feature extraction and robustness against lighting variations.
Trained on 4,503 labeled images with data augmentation techniques to enhance model generalization.
DenseNet201 + SVM model achieved 95% accuracy, outperforming standalone CNNs and other hybrids.
The model exhibited improved performance in challenging conditions like noise and lighting inconsistencies.
Comprehensive metrics such as precision, recall, and F1-score confirmed strong classification performance.

Abstract

Purpose: Early and accurate detection of plant leaf diseases is vital for safe-guarding crop yield and supporting sustainable agricultural practices. However,practical deployment faces challenges such as inconsistent lighting conditions,overlapping leaves, low-contrast early-stage symptoms, and noisy image data—allof which hinder the reliability of deep learning models in field environments.This study aims to develop a robust, scalable, and interpretable classificationframework capable of performing effectively under such real-world conditions.Method: We propose a hybrid classification pipeline that integrates deep featureextraction from three pre-trained Convolutional Neural Networks (CNNs) basedVGG16, ResNet50, and DenseNet201—with a linear Support Vector Machine(SVM) classifier. To enhance robustness to varying illumination, all imagesare converted from RGB to HSV colour space, enabling chromatic features tobe isolated from brightness fluctuations. Features are extracted from the finalglobal pooling layers of each CNN, then concatenated to construct a unifiedhigh-dimensional feature vector. This vector is passed to the SVM classifier forbinary classification (healthy vs. diseased). The system was trained and validatedusing a publicly available dataset comprising 4,503 labelled images, balancedbetween healthy and diseased samples. A comprehensive data augmentationstrategy—including rotation, flipping, and scaling—was employed to improvegeneralisation and mitigate overfitting.Results: Among the evaluated configurations, the DenseNet201 + SVM modelachieved the highest accuracy of 95%, outperforming both standalone CNN mod-els and other hybrid variants including, VGG16-SVM, and ResNet50-SVM. Thehybrid approach demonstrated enhanced generalisability, particularly in imagesaffected by noise and lighting inconsistencies. Precision, recall, F1-score, andconfusion matrix metrics confirmed the model’s strong performance across bothclasses.Conclusion: The proposed hybrid CNN-SVM framework offers a robust andinterpretable solution for real-world leaf disease detection. By leveraging HSVcolour space transformation and combining diverse CNN feature representations,the model effectively addresses common challenges in agricultural image classi-fication. This work presents a scalable pipeline with potential for deploymentin precision agriculture systems, including smartphone-based or drone-assistedmonitoring platforms.

Ask AI

Helpful

Bookmark

View Full Paper