LMP-PM: a lightweight multi-path pruning method for plant leaf disease recognition.

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Plant leaf diseases significantly impact crop yield and productivity.
• Accurate and timely disease identification is crucial for effective management.
• Existing high-performance deep learning models are often too complex for resource-constrained agricultural environments.

Purpose of the Study:

• To develop an efficient and accurate lightweight approach for plant disease identification.
• To address the limitations of complex deep learning models in real-world agricultural settings.
• To create a flexible method for optimizing model size and performance.

Main Methods:

• Developed the Lightweight Multi-Path Pruning Method (LMP-PM) for model optimization.
• Constructed an original complex model (OMNet) with a three-branch parallel module (TBP block).
• Applied LMP-PM to OMNet, generating lightweight models and identifying the optimal one (LMNet).

Main Results:

• LMNet utilizes only 5.69% of OMNet's parameters and 3.80% of its FLOPs.
• LMNet achieved 99.23% accuracy on the Plant Village dataset, outperforming OMNet by 0.58%.
• LMNet achieved 87.27% accuracy on the AI 2018 Challenger dataset, surpassing OMNet by 1.91%.

Conclusions:

• LMP-PM successfully creates highly efficient lightweight models like LMNet.
• LMNet drastically reduces computational resources and improves classification accuracy.
• LMNet is suitable for real-time plant disease identification in resource-constrained agricultural environments.