Paddy leaf disease detection and classification using improved Gorilla Troops optimized YOLO-V8 network

Paddy leaf disease (PLD) detection has grown more difficult, yet early detection might prevent significant losses due to decreased crop yield. However, existing models struggle to accurately classify diseases under difficult circumstances like intricate backgrounds, fluctuating lighting, and overlapping leaves. Additionally, existing models do not incorporate efficient optimization strategies, leading to suboptimal accuracy and poor generalization on unseen data. To address these challenges, a novel deep learning-based YOLO-LEAFNET method for PLD detection utilizing IGT-YOLO, integrating the YOLOv8 disease detection with the Improved Gorilla Troops (IGT) optimization. The input paddy leaf images are pre-processed using Bilateral Contrast Limited Adaptive Histogram Equalization (B-CLAHE) to enhance image quality and improve local contrast while preserving disease boundaries. YOLOv8 model is utilized to detect and classify paddy leaf diseases by accurately localizing affected regions with bounding boxes. Then, the IGT algorithm boosts the disease detection accuracy by optimizing YOLOv8 through effective hyperparameter tuning. The proposed YOLO-LEAFNET method effectiveness was evaluated using recall, F1 score, specificity, accuracy, and precision. B-CLAHE enhanced noise-free images improve contrast and detection accuracy, while the IGT-YOLO model ensures scalable, efficient early diagnosis of paddy leaf diseases with 99.07% accuracy. The YOLO-LEAFNET enhanced the total accuracy by 3.21%, 5.25%, and 1.98% related to CNN, DeepRice, and FasterR-CNN, respectively.