Research on grape leaf disease recognition method based on improved YOLOv8n model.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Grape leaf disease recognition models often suffer from large sizes and limited classification capabilities.
• Existing models struggle with efficiency and accuracy in real-world vineyard conditions.

Purpose of the Study:

• To develop an enhanced grape leaf disease recognition model with improved accuracy, reduced size, and faster inference speed.
• To address the limitations of current models for practical application in precision agriculture.

Main Methods:

• Modified the YOLOv8n architecture by incorporating G-bneck and simSPPF modules for enhanced feature extraction and speed.
• Introduced the C2F-UIB module to reduce model parameters and computational load.
• Integrated LInner-CIoU loss and partial convolution for accelerated inference and robust image handling.

Main Results:

• The enhanced YOLOv8n model achieved 97.3% accuracy, a 3.53MB model size, and 228.55 FPS.
• Outperformed YOLOv3-tiny, YOLOv5n, and YOLOv6n in average accuracy.
• Demonstrated 89.3% accuracy and 5.18 seconds processing time on a spraying device.

Conclusions:

• The proposed model effectively overcomes the challenges of size and speed in grape leaf disease recognition.
• Enables accurate and rapid identification of grape leaf diseases in natural environments.
• Offers significant potential for developing effective grape disease management technologies in precision agriculture.