PLDC-Net: A Domain-Specific Base Model for Plant Leaf Disease Classification Domain Adaptation Tasks

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Plant diseases cause significant crop losses and contribute to food shortages.
• Traditional disease monitoring is labor-intensive, costly, and prone to errors.
• Accurate and early disease identification is crucial for mitigating yield loss.

Purpose of the Study:

• To develop an improved deep learning model for plant leaf disease classification.
• To address the limitations of current domain adaptation methods in plant disease detection.
• To create a domain-specific pretraining strategy for enhanced model generalization.

Main Methods:

• Proposed PLDC-Net: an attention-based, SiLU-activated DenseNet201 architecture.
• Pretrained PLDC-Net on a large-scale, custom-built plant leaf disease dataset.
• Validated domain adaptation using transfer learning, fine-tuning, one-shot, and few-shot learning.

Main Results:

• Achieved over 24% improvement in F1-Score compared to baseline methods in domain adaptation.
• Demonstrated the effectiveness of domain-specific pretraining for plant disease classification.
• PLDC-Net shows superior performance in adapting to new plant and disease types.

Conclusions:

• Domain-specific pretraining with PLDC-Net offers a significant advancement over unrelated datasets like ImageNet.
• The proposed model enhances the accuracy and efficiency of plant disease identification.
• This research provides a robust foundation for AI-driven agricultural disease management.