Automatic Plant Disease Detection Based on Tranvolution Detection Network With GAN Modules Using Leaf Images.

Area of Science:

• Agricultural Science
• Computer Vision
• Artificial Intelligence

Background:

• Accurate plant disease detection is crucial for agricultural productivity and crop yield.
• Traditional deep learning methods face challenges including high hardware/data demands, slow inference, and poor generalization.
• Image variations due to lighting and complex leaf structures hinder disease detection.

Purpose of the Study:

• To develop an efficient and accurate plant disease detection system for practical agricultural use.
• To overcome the limitations of existing deep learning models in terms of speed, data requirements, and generalization.
• To propose a novel Tranvolution detection network integrated with Generative Adversarial Network (GAN) modules.

Main Methods:

• A Tranvolution architecture combining Transformer and Convolutional Neural Network (CNN) was developed.
• Generative models were incorporated into the backbone and attention extraction modules as GAN modules.
• The performance of various generative model combinations was validated, including SAGAN and WGAN.

Main Results:

• The proposed Tranvolution network achieved 51.7% Precision, 48.1% Recall, and 50.3% mAP.
• SAGAN demonstrated superior performance in the attention extraction module.
• WGAN proved most effective for image augmentation.

Conclusions:

• The novel Tranvolution network with GAN modules offers a promising solution for plant disease detection.
• The model's performance metrics indicate its potential for practical agricultural deployment.
• Deployment on an intelligent agricultural robot highlights the system's real-world applicability.