Multi-convolutional neural networks for cotton disease detection using synergistic deep learning paradigm.

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Cotton is a vital cash crop threatened by diseases, impacting global agriculture.
• Accurate disease detection is crucial for yield preservation and resource management.
• Existing automated methods face challenges like class imbalance and symptom variability.

Purpose of the Study:

• To develop an automated, accurate method for classifying seven cotton plant conditions (six diseases and healthy).
• To address challenges in multi-disease classification, including class imbalance and real-time detection needs.

Main Methods:

• Synthetic data generation using conventional techniques and a customized StyleGAN to address class imbalance.
• Feature extraction using MobileNet and VGG16, combined into a comprehensive feature vector.
• An ensemble classifier (StackNet) integrating Long Short Term Memory Units, Support Vector Machines, and Random Forest outputs.

Main Results:

• Achieved an average accuracy of 97% in classifying six cotton diseases and a healthy class.
• The proposed method demonstrated superior performance compared to existing state-of-the-art techniques.
• Successfully addressed class imbalance and improved classification accuracy for diverse disease symptoms.

Conclusions:

• The novel deep learning ensemble method offers a robust solution for accurate, multi-class cotton disease identification.
• This approach has the potential to significantly aid farmers in disease management and improve crop yields.
• The method's high accuracy and ability to handle real-world challenges pave the way for practical agricultural applications.