Classification of cotton leaf disease using YOLOv8 based k-fold cross validation deep learning method for precision...

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Cotton is a vital global crop, supporting millions of farmers and the textile industry.
• Crop yield and quality are significantly impacted by diseases and environmental stresses.
• Deep learning offers promising solutions for early disease detection and management in agriculture.

Purpose of the Study:

• To develop and validate a precise method for identifying multiple cotton leaf diseases.
• To leverage the YOLOv8 deep learning model combined with 10-fold cross-validation for robust disease recognition.
• To improve cotton disease management strategies through advanced computational techniques.

Main Methods:

• Utilized the YOLOv8 deep learning model for image classification of cotton leaf diseases.
• Implemented a 10-fold cross-validation technique to ensure model generalizability and mitigate overfitting.
• Trained and tested the model using field-captured images of cotton plants.

Main Results:

• Achieved high performance metrics, including Top-1 accuracy of 99.60% and Top-5 accuracy of 100% in initial testing.
• Demonstrated consistent performance across 10 trials, with average Top-1 accuracy of 98.41% and average recall, precision, and F1 score above 98%.
• The model showed excellent precision (99.53%), recall (99.53%), and F1 score (99.60%) for disease identification.

Conclusions:

• The proposed YOLOv8 model with 10-fold cross-validation is a highly effective tool for multi-class cotton leaf disease identification.
• This approach offers a reliable and accurate solution for early disease detection, potentially increasing cotton yield and quality.
• This study pioneers the application of YOLOv8 classification with k-fold cross-validation for identifying cotton diseases from field images.