RTCB: an integrated deep learning model for garlic leaf disease identification.

Area of Science:

• Agricultural Science
• Computer Vision
• Deep Learning

Background:

• Garlic cultivation is vital, but leaf diseases and pests significantly reduce crop yield.
• Traditional disease detection methods are time-consuming and lack accuracy.
• Accurate and efficient disease identification is crucial for intelligent agriculture.

Purpose of the Study:

• To develop an advanced deep learning model for accurate garlic leaf disease recognition.
• To improve computational efficiency and feature extraction in disease detection.
• To provide a scalable solution for automated disease monitoring in agriculture.

Main Methods:

• An upgraded ResNet18 architecture was employed, incorporating partial convolutions for efficiency.
• A triplet attention mechanism was introduced to enhance focus on critical disease features.
• A convolutional block attention mechanism was added to each residual layer for improved feature perception.

Main Results:

• The proposed ResNet18, triplet, convolutional block (RTCB) attention model achieved 98.90% classification accuracy.
• The RTCB model outperformed other leading deep learning models in accuracy and speed.
• The model demonstrated superior recognition precision and generalization ability.

Conclusions:

• The RTCB model offers a highly accurate and efficient solution for garlic leaf disease detection.
• This approach provides a valuable technical reference for automated disease monitoring in intelligent agriculture.
• The model's efficiency supports deployment on edge computing devices for practical applications.