Related Experiment Video
Updated: Jan 18, 2026
06:41
Imaging and Analysis for Quantifying Maize (Zea mays) Abiotic Stress Phenotypes
Published on: March 28, 2025
1.6K
Sparse-MoE-SAM: A Lightweight Framework Integrating MoE and SAM with a Sparse Attention Mechanism for Plant Disease
Benhan Zhao1, Xilin Kang2, Hao Zhou1
1School of Electronic Information and Physics, Central South University of Forestry and Technology, Changsha 410004, China.
Plants (Basel, Switzerland)
|September 13, 2025
Summary
This study introduces Sparse-MoE-SAM, an efficient AI model for plant disease segmentation. It achieves high accuracy on limited hardware by using sparse attention and a mixture of experts (MoE) decoder, significantly reducing computational costs.
Area of Science:
- Artificial Intelligence
- Computer Vision
- Plant Pathology
Background:
- Deploying AI for plant disease segmentation is hindered by computational demands of attention mechanisms, challenges in capturing sparse lesion details, and segmentation errors from complex field conditions.
- Traditional dense attention models exhibit quadratic complexity, making them unsuitable for resource-constrained environments.
- Accurate segmentation requires models that can handle both long-range dependencies and fine local details, especially with varying lesion sizes and complex backgrounds.
Purpose of the Study:
- To develop an efficient and accurate plant disease segmentation framework for resource-limited settings.
- To address the computational complexity and segmentation accuracy challenges in current AI models for plant disease identification.
- To enable the deployment of advanced segmentation models on edge devices for real-time agricultural monitoring.
Main Methods:
- Proposed Sparse-MoE-SAM framework, integrating sparse attention mechanisms with a two-stage Mixture of Experts (MoE) decoder.
- Implemented sparse attention to dynamically activate channels, reducing complexity while maintaining context.
- Designed a sparse attention-enhanced Atrous Spatial Pyramid Pooling (ASPP) module for multi-scale feature extraction.
Main Results:
- Sparse-MoE-SAM achieved a mean Intersection-over-Union (mIoU) of 94.2% on diverse datasets, outperforming standard Segment Anything Model (SAM) by 2.5%.
- Reduced computational costs by 23.7% compared to the original SAM baseline.
- Demonstrated robust performance across different plant textures, lesion morphologies, and lighting conditions, with enhanced hardware compatibility.
Conclusions:
- Integrating sparse attention with MoE mechanisms effectively reduces computational demands while maintaining high segmentation accuracy.
- The Sparse-MoE-SAM framework offers a viable solution for scalable deployment of plant disease segmentation on mobile and edge devices.
- This approach enhances the practical applicability of AI in precision agriculture for early disease detection and management.
Related Concept Videos
02:16Light Acquisition
9.4K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
9.4K
02:35Key Elements for Plant Nutrition
24.0K
Like all living organisms, plants require organic and inorganic nutrients to survive, reproduce, grow and maintain homeostasis. To identify nutrients that are essential for plant functioning, researchers have leveraged a technique called hydroponics. In hydroponic culture systems, plants are grown—without soil—in water-based solutions containing nutrients. At least 17 nutrients have been identified as essential elements required by plants. Plants acquire these elements from the...
24.0K