Related Experiment Video
Updated: Jun 14, 2025
06:41
Imaging and Analysis for Quantifying Maize (Zea mays) Abiotic Stress Phenotypes
Published on: March 28, 2025
731
MnasNet-SimAM: An Improved Deep Learning Model for the Identification of Common Wheat Diseases in Complex Real-Field
Xiaojie Wen1,2, Muzaipaer Maimaiti1,2, Qi Liu1,2
1Key Laboratory of the Pest Monitoring and Safety Control of Crops and Forests of the Xinjiang Uygur Autonomous Region, College of Agronomy, Xinjiang Agricultural University, Urumqi 830052, China.
Plants (Basel, Switzerland)
|August 29, 2024
Summary
This study introduces MnasNet-SimAM, a novel deep learning model for accurate wheat disease detection in complex natural settings. The model achieves high accuracy, improving upon existing methods for agricultural disease identification.
Area of Science:
- Agricultural Science
- Computer Vision
- Deep Learning
Background:
- Deep learning is crucial for agricultural disease detection but faces challenges with complex backgrounds and similar disease identification.
- Accurate detection of wheat diseases is vital for crop management and yield optimization.
Purpose of the Study:
- To develop and evaluate a deep learning model for detecting six prevalent wheat diseases and healthy wheat in complex natural environments.
- To address limitations in recognition accuracy and misjudgment rates of existing agricultural disease detection methods.
Main Methods:
- Evaluated five lightweight convolutional neural networks for wheat disease recognition.
- Developed a novel model, MnasNet-SimAM, by integrating transfer learning with an attention mechanism (SimAM).
- Tested model performance on images captured in complex natural contexts and a public dataset.
Main Results:
- Five lightweight convolutional neural networks achieved over 90% accuracy in recognizing six wheat diseases.
- MnasNet-SimAM reached 95.14% accuracy, a 1.7% improvement over the base model with minimal parameter increase.
- MnasNet-SimAM demonstrated strong generalization with 91.20% accuracy on the Wheat Fungi Diseases dataset.
Conclusions:
- The MnasNet-SimAM model effectively addresses challenges in agricultural disease detection, particularly for wheat.
- The proposed model meets the need for rapid and accurate wheat disease identification in real-world conditions.
- This research contributes to advancing deep learning applications in precision agriculture.
