DCNet: DenseNet-77-based CornerNet model for the tomato plant leaf disease detection and classification

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Tomato leaf disease identification is complex due to visual similarities between healthy and diseased areas.
• Image variations like lighting, color, and noise further complicate accurate disease detection.
• Manual disease identification methods are labor-intensive and prone to errors.

Purpose of the Study:

• To develop a robust deep learning approach for early tomato plant leaf disease detection and classification.
• To address the challenges posed by image variations and similarities in disease identification.
• To provide an automated system for agriculturalists to improve crop yield and reduce costs.

Main Methods:

• Proposed a novel DenseNet-77-based CornerNet model for disease localization and classification.
• Utilized DenseNet-77 as the backbone for feature extraction from tomato leaf images.
• Employed a one-stage detector (CornerNet) to classify abnormalities into 10 distinct disease classes.

Main Results:

• Achieved an average accuracy of 99.98% on the challenging PlantVillage dataset.
• The model demonstrated effectiveness in handling variations in brightness, color, and image dimensions.
• Successfully localized and classified tomato plant leaf abnormalities with high precision.

Conclusions:

• The DenseNet-77-based CornerNet model offers a highly accurate and efficient solution for tomato leaf disease detection.
• This automated approach can significantly assist agriculturalists in timely disease management, improving crop yield.
• The proposed method provides a robust alternative to traditional manual disease identification systems.