Comparative Analysis of Deep Convolution Neural Networks for Classification of Fungal Diseases in Guava Leaves
Classification of Fungal Diseases in Guava Leaves
Keywords:
Guava Leaves Classification , Deep Learning , Fungal Disease , Multiclass Classification , Fine TuningAbstract
Guava, nutrients-rich fruit, is severely affected by various fungal leaf diseases, which hinder fruit development, optimal growth, and bud formation. Among the 177 pathogens known to damage guava leaves, 167 are fungal. However, Effective classification of these diseases is crucial to minimizing economic losses in the food industry. This study focuses on classifying common six fungal diseases in guava leaves, including Pseudocercospora leaf spot, Anthracnose, Canker, Rust, Mummification, and Dot. Deep learning models, such as DenseNet (169,121), EfficientNet (B3), squeeznet, and Alexnet have been used for disease classification but suffer from overfitting and low accuracy on this dataset. To address these limitations, this study evaluates finetuned CNN architectures: EfficientNetB2, EfficientNetB4, and DenseNet201 using both existing and real-world datasets. The models are assessed based on accuracy, precision, recall, and F1 score to determine their effectiveness in classifying fungal diseases. The findings will provide insights into the performance of deep learning algorithms in guava leaf disease classification and contribute to improved agricultural disease management.
