Related Experiment Video
Updated: Jun 20, 2026
08:47
Computer Vision-Based Biomass Estimation for Invasive Plants
Published on: February 9, 2024
Multi-scale feature fusion-based vision mamba for robust plant disease image classification on field-acquired
Shanjiang Zhang1,2, Renjing Liu1
1School of Management, Xi'an Jiaotong University, Xi'an, China.
Frontiers in Plant Science
|June 19, 2026
Summary
A new Vision Mamba model enhances plant disease identification by effectively fusing multi-scale features and using attention mechanisms. This lightweight approach improves accuracy in complex field conditions for precision agriculture.
Area of Science:
- Agricultural Computer Vision
- Deep Learning for Plant Pathology
Background:
- Existing deep learning models struggle to simultaneously capture local lesion details and global context in field plant images.
- There is a need for efficient, lightweight models for plant disease identification in real-world agricultural settings.
Purpose of the Study:
- To design and evaluate an improved Vision Mamba network for accurate plant disease classification.
- To address limitations in capturing fine-grained local features and long-range dependencies in plant disease images.
Main Methods:
- An improved Vision Mamba network incorporating a Multi-Scale Feature Fusion Module (MFFM), Adaptive Channel Attention Mechanism (ACAM), and Lightweight Residual Connection (LRC).
- MFFM fuses features from different network depths; ACAM focuses on relevant channels and reduces background noise; LRC mitigates gradient issues in deep networks.
- The model was evaluated on the PlantDoc dataset.
Main Results:
- The proposed model achieved 92.67% overall accuracy, outperforming the baseline Vision Mamba by 5.33%.
- Stable accuracy of 92.41 ± 0.24% was observed through five-fold cross-validation, with statistically significant improvements (p<0.05).
- Ablation studies confirmed the effectiveness of the integrated MFFM, ACAM, and LRC modules.
Conclusions:
- The improved Vision Mamba network demonstrates significant potential for plant disease identification in agricultural computer vision.
- The model's efficiency and accuracy make it suitable for edge deployment in precision agriculture and intelligent crop disease management.