An automatic plant leaf disease identification using DenseNet-121 architecture with a mutation-based henry gas solubility optimization algorithm
Farmers are struggling to provide the fast-growing population with sufficient agricultural products, while plant diseases result in devastating food loss. The billions of dollars spent by agriculturists in disease management often result in poor disease control without any technical support. Advance...
Gespeichert in:
Veröffentlicht in: | Neural computing & applications 2022-04, Vol.34 (7), p.5513-5534 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | Farmers are struggling to provide the fast-growing population with sufficient agricultural products, while plant diseases result in devastating food loss. The billions of dollars spent by agriculturists in disease management often result in poor disease control without any technical support. Advances in computer vision techniques help to detect plant pathogens at an earlier level with an adaptive algorithm designed through deep learning and machine learning techniques. In this paper, we present an efficient Mutation-based Henry Gas Solubility Optimization (MHGSO) algorithm to optimize the hyperparameters of the DenseNet-121 architecture. The hyperparameter optimization is mainly done to reduce the computational complexity and the error rate of the Convolutional Neural Network (CNN). This step helps the MHGSO optimized DenseNet-121 architecture to achieve a higher classification accuracy for classifying different plant images from the PlantVillage dataset. The experimental results achieved showed that the proposed model is capable of classifying 14 leaf classes present in the PlantVillage dataset with higher classification accuracy (98.7%) and stability. When tested with a field dataset with complicated backgrounds, the proposed MHGSO optimized DenseNet-121 architecture achieves accuracy, precision, and recall scores of 98.81%, 98.60%, and 98.75%, respectively. |
---|---|
ISSN: | 0941-0643 1433-3058 |
DOI: | 10.1007/s00521-021-06714-z |