CapPlant: a capsule network based framework for plant disease classification.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate plant disease classification is crucial for understanding plant health and optimizing agricultural practices.
• Image-based disease recognition presents significant challenges in agriculture due to complex visual variations.

Purpose of the Study:

• To propose a novel deep learning architecture, CapPlant, for automated plant disease classification from images.
• To leverage capsule layers for improved feature extraction, capturing spatial relationships for precise disease prediction.

Main Methods:

• Developed the CapPlant deep learning model incorporating convolutional layers and a capsule layer.
• Utilized the PlantVillage dataset, comprising over 50,000 images of healthy and diseased plants, for model training and evaluation.
• Compared CapPlant's performance against existing plant disease classification models.

Main Results:

• The CapPlant model achieved a high overall test accuracy of 93.01%.
• The model demonstrated a strong F1 score of 93.07%, indicating effective disease classification.
• CapPlant outperformed other models in prediction accuracy for plant disease identification.

Conclusions:

• The CapPlant architecture offers a significant improvement in automated plant disease classification accuracy.
• The integration of capsule layers enhances the model's ability to understand spatial relationships for precise diagnosis.
• This deep learning approach holds promise for advancing agricultural monitoring and disease management strategies.