AI-driven plant disease detection with tailored convolutional neural network.

Area of Science:

• Agricultural technology
• Artificial intelligence in agriculture
• Machine learning for crop monitoring

Background:

• Deep learning models are increasingly used in agriculture for tasks like disease identification and yield prediction.
• A significant challenge lies in developing efficient, lightweight, and cost-effective deep learning models suitable for deployment on resource-constrained devices.

Purpose of the Study:

• To address the need for efficient deep learning models in agriculture.
• To propose a Convolutional Neural Network (CNN) architecture optimized by a Genetic Algorithm (GA) for automated hyperparameter selection.
• To ensure high performance with minimal computational overhead for deployment on small devices.

Main Methods:

• Developed a novel Convolutional Neural Network (CNN) architecture.
• Utilized a Genetic Algorithm (GA) to automate the selection of critical hyperparameters (e.g., number and size of filters).
• Created a custom dataset of tea leaf diseases, including pest-induced and pathogen-induced conditions.

Main Results:

• The proposed GA-based CNN achieved 97.6% accuracy on the custom tea leaf disease dataset.
• The model demonstrated robustness on external datasets, reaching 96.99% on PlantVillage and 99% on a Rice leaf disease dataset.
• Outperformed several state-of-the-art deep learning models in terms of accuracy and parameter efficiency.

Conclusions:

• The GA-optimized CNN offers an efficient and accurate solution for crop disease identification.
• The model's lightweight design makes it suitable for deployment on edge devices in agricultural settings.
• This approach successfully automates hyperparameter optimization, reducing computational costs and improving model performance.