An Instance Segmentation Model for Strawberry Diseases Based on Mask R-CNN.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate plant disease identification is vital for minimizing economic and quality losses in agriculture.
• Deep neural networks offer advanced solutions for agricultural challenges, but limited datasets hinder autonomous strawberry disease detection.
• Fine-grained instance segmentation for strawberry diseases requires specialized datasets and robust models.

Purpose of the Study:

• To introduce a novel, comprehensive dataset for strawberry disease detection.
• To develop and evaluate an automated system for fine-grained instance segmentation of strawberry diseases.
• To establish a baseline for deep learning-based strawberry disease detection systems.

Main Methods:

• A new dataset of 2500 images featuring seven types of strawberry diseases was created.
• A Mask R-CNN architecture with a ResNet backbone was employed for instance segmentation.
• Systematic data augmentation techniques were applied to enhance model performance under complex conditions.

Main Results:

• The proposed model achieved a mean average precision (mAP) of 82.43% for strawberry disease instance segmentation.
• The model demonstrated effective segmentation of seven distinct strawberry diseases.
• Performance was validated under complex environmental and background conditions.

Conclusions:

• The developed dataset and Mask R-CNN model provide a strong foundation for autonomous strawberry disease detection.
• The approach offers a low-cost, highly accurate solution for early disease identification in strawberries.
• This work facilitates the advancement of deep learning applications in precision agriculture.