Identification of rice leaf disease based on DepMulti-Net

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Rice leaf disease identification faces challenges like complex backgrounds and feature extraction.
• Existing models often have large parameter volumes, limiting efficiency.

Purpose of the Study:

• To develop a novel, efficient, and lightweight model for rice leaf disease identification.
• To address challenges in feature extraction and complex backgrounds in disease detection.

Main Methods:

• Constructed a dataset of 20,000 rice disease images covering four common diseases.
• Introduced a VGG-block module with depth-separable convolution to reduce parameters.
• Designed a multi-scale feature fusion module and integrated feature reuse with an inverse bottleneck structure.

Main Results:

• DepMulti-Net achieved 98.56% average accuracy in identifying four major rice diseases.
• The model has a significantly reduced parameter count of only 13.50M.
• Outperformed existing rice leaf disease identification methods in experimental evaluations.

Conclusions:

• DepMulti-Net provides an efficient and lightweight solution for crop disease identification.
• The model's design effectively handles complex backgrounds and enhances fine-grained feature recognition.
• This research contributes to the advancement of smart agriculture through improved disease detection technology.