Related Experiment Video
Updated: Mar 15, 2026
05:03
Author Spotlight: Advancing Stomatal Research with Automated Aperture Measurement
Published on: February 9, 2024
2.4K
MSP-Net: An Effective Multi-Scale Feature-Aware Detection Network for the Detection of Tomato Leaf Diseases
Feng Kang1,2,3, Lijin Wang1,3,4, Huicheng Li1,3
1College of Computer and Information Sciences, Fujian Agriculture and Forestry University, Fuzhou 350002, China.
Plants (Basel, Switzerland)
|March 14, 2026
Summary
This study introduces MSP-Net for accurate automatic tomato leaf disease detection, improving precision agriculture. The lightweight L-MSP-Net variant is optimized for edge devices, enhancing efficiency in real-world applications.
Area of Science:
- Agricultural Science
- Computer Vision
- Machine Learning
Background:
- Precision agriculture requires robust automated systems for crop disease detection.
- Challenges include variable lesion appearance, environmental noise, and computational constraints for field deployment.
Purpose of the Study:
- To develop an advanced automatic tomato leaf disease detection framework.
- To address limitations of existing methods in complex field conditions.
- To create an efficient, deployable model for edge computing.
Main Methods:
- Proposed MSP-Net framework with Multi-Scale Perception Convolution Module (MSPCM), SimAM-enhanced C3k2 layers, and Multi-Scale Feature Enhancement Module (MSFEM).
- Developed lightweight L-MSP-Net using architectural migration and structured pruning for edge efficiency.
- Evaluated on the Tomato-Village dataset and cross-dataset experiments on PASCAL VOC and MS COCO.
Main Results:
- MSP-Net achieved 92.0% mAP@0.5, outperforming YOLOv11s by 2.0%.
- L-MSP-Net attained 86.1% mAP@0.5, improving by 3.6% over YOLOv11n while reducing parameters by 10.5%.
- Successful deployment of L-MSP-Net on the RK3588 edge platform.
Conclusions:
- The proposed MSP-Net framework significantly enhances tomato leaf disease detection accuracy.
- The lightweight L-MSP-Net offers efficient edge deployment for precision agriculture.
- Architectural refinements demonstrate transferability to general object detection tasks.
