GAPNet: Single and multiplant leaf disease classification method based on simplified SqueezeNet for grape, apple and...

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Agriculture is vital for national economies and food security.
• Early detection of plant diseases is critical for crop yield and quality.
• Existing methods for plant disease classification are being improved with AI.

Purpose of the Study:

• To develop a simple, effective, and lightweight model for plant leaf disease classification.
• To evaluate the performance of various pre-trained convolutional neural network (CNN) architectures.
• To propose a novel simplified CNN model, GAPNet, for enhanced disease identification.

Main Methods:

• Utilized seven state-of-the-art pre-trained CNNs (VGG16, ResNet50, SqueezeNet, Xception, ShuffleNet, DenseNet121, MobileNetV2).
• Developed GAPNet, a simplified SqueezeNet-based model optimized for speed and efficiency.
• Applied synthetic minority oversampling technique (SMOTE) to address data imbalance.
• Tested the model on grape, apple, and potato leaf disease datasets.

Main Results:

• GAPNet achieved high accuracy: 99.72% (grape), 99.53% (apple), and 99.83% (potato).
• A combined multi-plant classification accuracy of 99.64% was recorded.
• GAPNet demonstrated superior performance compared to other state-of-the-art methods.

Conclusions:

• The lightweight GAPNet model offers a promising solution for accurate and efficient plant disease classification.
• This approach can significantly contribute to early disease diagnosis, improving agricultural output.
• The model's effectiveness across multiple plant species highlights its broad applicability in agriculture.