Automatic prediction of cotton leaf's diseases using deep learning techniques.

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Cotton leaf diseases significantly reduce global yield and fiber quality.
• Traditional disease diagnosis is inefficient for large-scale farming.
• Existing deep learning models struggle with real-world field data and preprocessing.

Purpose of the Study:

• To develop an optimized transfer learning model for efficient cotton leaf disease identification.
• To improve the generalization and accuracy of deep learning models in field conditions.
• To create a practical tool for farmers to monitor cotton plant health.

Main Methods:

• Developed the Cotton Leaf Disease Prediction Convolutional Neural Network (CLDP-CNN) using Transfer Learning (TL).
• Trained the model on two distinct datasets: field images and Kaggle-sourced images.
• Evaluated model performance on real-world field datasets, including VGG16 pre-trained models.

Main Results:

• CLDP-CNN achieved 99.78% accuracy on the primary field dataset and 99.62% on the secondary dataset.
• The VGG16 model attained 99.56% accuracy on the primary dataset and 98.82% on the secondary dataset.
• High detection success rates demonstrate the model's effectiveness in identifying cotton leaf diseases.

Conclusions:

• The CLDP-CNN model offers a highly accurate and efficient solution for cotton leaf disease detection.
• Transfer learning significantly enhances model performance and generalization for agricultural applications.
• A web-based application provides real-time insights, empowering farmers with timely interventions to prevent crop loss.