Related Experiment Video
Updated: Sep 5, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
1.6K
A Vegetable Leaf Disease Identification Model Based on Image-Text Cross-Modal Feature Fusion.
Xuguang Feng1,2,3,4, Chunjiang Zhao2,3, Chunshan Wang1,2,3,4
1School of Information Science and Technology, Hebei Agricultural University, Baoding, China.
Frontiers in Plant Science
|July 11, 2022
Summary
This study introduces a novel YOLOv5s + Bidirectional Cross-Modal Transformer (BiCMT) model for accurate crop disease identification in complex field conditions. The model effectively fuses image and text data, significantly improving disease detection accuracy and robustness.
Area of Science:
- Agricultural Science
- Computer Vision
- Artificial Intelligence
Background:
- Automatic crop disease identification is crucial for smart agriculture but challenging due to complex backgrounds and varied disease appearances.
- Existing Deep Convolutional Neural Network (DCNN) models struggle with accuracy and robustness in complex field environments.
Purpose of the Study:
- To develop an end-to-end model for accurate and robust field disease identification.
- To enhance disease identification by integrating regional attention and cross-modal feature fusion.
Main Methods:
- Proposed a hybrid model combining YOLOv5s for disease-spot region detection and Bidirectional Cross-Modal Transformer (BiCMT) for classification.
- YOLOv5s provides regional attention, guiding the BiCMT classifier.
- BiCMT fuses image and text data, addressing sequence length inconsistencies for comprehensive feature recognition.
Main Results:
- The YOLOv5s + BiCMT model achieved optimal results on a small dataset.
- Achieved high performance metrics: 99.23% Accuracy, 97.37% Precision, 97.54% Sensitivity, and 99.54% Specificity.
- Demonstrated the effectiveness of bidirectional cross-modal feature fusion for vegetable disease identification.
Conclusions:
- Bidirectional cross-modal feature fusion using image and text data is an effective strategy for field vegetable disease identification.
- The proposed YOLOv5s + BiCMT model offers a robust solution for smart agriculture applications.
- This approach enhances disease detection accuracy in challenging agricultural settings.
