Attention-enhanced corn disease diagnosis using few-shot learning and VGG16

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Traditional plant disease detection is manual, time-consuming, and labor-intensive.
• AI, Machine Learning, and Deep Learning methods require extensive datasets, which are difficult to obtain and annotate.
• Few-Shot Learning (FSL) offers a solution by enabling model generalization with minimal training examples, mimicking human learning.

Purpose of the Study:

• To develop an efficient and accurate AI model for early-stage plant disease detection.
• To address the data acquisition and annotation challenges in traditional machine learning approaches for plant pathology.
• To provide a feasible solution for real-world agricultural applications through Few-Shot Learning.

Main Methods:

• Utilized a pre-trained VGG16 convolutional neural network as the backbone.
• Integrated an attention module with the VGG16 backbone.
• Applied prototypical Few-Shot Learning for corn disease prediction and classification.

Main Results:

• Achieved a high accuracy of 98.25% in corn disease prediction and classification.
• Demonstrated the effectiveness of Few-Shot Learning in overcoming data limitations.
• Developed a robust and scalable solution for agricultural disease identification.

Conclusions:

• The proposed Few-Shot Learning model enables early and accurate identification of corn diseases.
• The system's feasibility is enhanced by overcoming the need for large annotated datasets.
• The integration of VGG16, attention mechanisms, and prototypical FSL offers a powerful tool for farmers to mitigate crop losses.