Developing sustainable system based on transformers algorithms to predict the Dubas insects diseases in palm leaves

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Early identification of plant diseases is vital for mitigating crop yield loss.
• Palm tree health, particularly leaf condition, directly impacts agricultural productivity.
• Pests and diseases pose a significant threat to palm cultivation, reducing overall yield.

Purpose of the Study:

• To develop and evaluate computer vision models for detecting diseases and pests in palm leaves.
• To enhance pest management strategies in agriculture through advanced image analysis.
• To compare the efficacy of EfficientNetV2B0, DenseNet12, and Vision Transformer (ViT) models for palm leaf disease identification.

Main Methods:

• Utilized a dataset of 1600 palm leaf images (800 healthy, 800 diseased).
• Developed and trained EfficientNetV2B0, DenseNet12, and Vision Transformer (ViT) models.
• Employed image analysis techniques for disease and pest detection.

Main Results:

• The Vision Transformer (ViT) model achieved a high accuracy of 99.37%.
• Proposed models outperformed several recent studies in palm leaf disease identification.
• Performance was validated on both original and augmented datasets.

Conclusions:

• The study presents an innovative approach for identifying palm leaf diseases.
• The developed models show significant promise for implementation in commercial agriculture.
• Effective pest and disease management can be improved through these advanced computational methods.