Apple Leaf Diseases Recognition Based on An Improved Convolutional Neural Network.

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Accurate identification of apple leaf diseases like scab, frogeye spot, and cedar rust is crucial for apple production.
• Existing diagnostic methods may lack speed and accuracy, hindering effective disease management.

Purpose of the Study:

• To develop an improved deep convolutional neural network model for rapid and accurate identification of common apple leaf diseases.
• To enhance diagnostic efficiency in apple production through advanced image recognition.

Main Methods:

• An improved VGG16 model was developed, incorporating a global average pooling layer and a batch normalization layer.
• Transfer learning was employed to reduce model training time and improve convergence speed.
• The model was trained and evaluated for apple leaf disease classification.

Main Results:

• The proposed model achieved an overall accuracy of 99.01% in classifying apple leaf diseases.
• Model parameters were reduced by 89% compared to the classical VGG16.
• Recognition accuracy increased by 6.3%, and training time was reduced to 0.56% of the original model's duration.

Conclusions:

• The improved VGG16-based deep convolutional neural network offers a highly accurate and efficient solution for identifying apple leaf diseases.
• The model's reduced parameters and faster convergence speed make it a practical tool for disease diagnosis in agriculture.
• This approach supports sustainable apple production through timely and precise disease management.