Optimizing Plant Leaf Disease Detection: A Comparative Study of Deep Learning Models
Main Article Content
Abstract
This study presents an advanced plant leaf disease detection system utilizing deep learning techniques to address the limitations of traditional disease identification methods. We evaluated several machine learning models, including Convolutional Neural Networks (CNN), VGG16, ResNet50, and EfficientNet-B2, using a dataset of over 87,000 RGB images of healthy and diseased crop leaves. Among these, EfficientNet-B2 demonstrated superior performance, achieving an accuracy of 93%, precision of 93%, recall of 90%, and an F1-score of 90%. The model's high sensitivity of 99.83% underscores its efficacy in identifying positive cases accurately. Despite signs of overfitting, as indicated by the divergence between training and validation metrics, EfficientNet-B2 remains the most reliable model for this task. Future work will focus on mitigating overfitting through techniques like hyperparameter tuning and data augmentation. Integrating this model into mobile and UAV-based platforms and expanding the dataset to include diverse plant species will enhance its practical application in real-world agricultural settings..