An Enhanced Deep Learning Model for Effective Crop Pest and Disease Detection

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Traditional machine learning methods face challenges in plant pest and disease image recognition due to small sample sizes and indistinct features.
• Accurate identification of crop pests and diseases is crucial for effective agricultural management and food security.

Purpose of the Study:

• To develop an improved deep learning model for enhanced crop pest and disease detection.
• To address limitations of existing methods in handling complex image recognition tasks in agriculture.

Main Methods:

• An improved ResNet34 model, termed ESA-ResNet34, was proposed, incorporating an efficient spatial attention (ESA) mechanism.
• Depthwise separable convolutions were utilized to reduce model parameters and computational load, alongside Dropout for overfitting mitigation.
• Data augmentation techniques including center cropping and horizontal flipping were applied to improve model robustness.

Main Results:

• The ESA-ResNet34 model achieved high performance metrics: 87.09% accuracy, 87.14% precision, and 86.91% F1 score.
• The proposed model significantly outperformed established benchmark models like AlexNet, VGG16, MobileNet, DenseNet, and other ResNet variants.
• Significant reductions in parameter count (85.37%) and computational load (84.51%) were achieved.

Conclusions:

• The ESA-ResNet34 model demonstrates superior performance in crop pest and disease detection compared to existing methods.
• The integration of efficient spatial attention and optimized convolutional layers offers a promising approach for agricultural image analysis.
• This advancement holds potential for improving precision agriculture and crop management strategies.