Related Experiment Video
Updated: Aug 5, 2026
06:22
A Precise and Autonomous System for the Detection of Insect Emergence Patterns
Published on: January 9, 2019
GrainPest-SSL: A Lightweight Semi-Supervised Detector for Stored-Grain Pest Monitoring in Smart Granaries
Yanbo Chen1, Xusheng Wei1, Huanran Wei1
1School of Computer Science and Artificial Intelligence, Nanjing University of Finance and Economics, Nanjing 210023, China.
Sensors (Basel, Switzerland)
|July 28, 2026
Summary
This study introduces GrainPest-SSL, an efficient framework for detecting stored-grain pests using limited data and resources. It enhances smart granary monitoring with improved pest identification accuracy and real-time processing capabilities.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Machine Learning
Background:
- Stored-grain pest monitoring is crucial for smart granaries but faces challenges with small, dense pests, high annotation costs, and limited edge computing.
- Existing methods struggle with accurate detection in complex backgrounds and resource-constrained environments.
Purpose of the Study:
- To develop an integrated framework (GrainPest-SSL) addressing bottlenecks in stored-grain pest monitoring.
- To improve the accuracy and efficiency of pest detection using limited labeled data and edge devices.
Main Methods:
- Construction of the GrainPest dataset with 1000 field images and 21,676 annotated pest instances.
- Design of a lightweight YOLOv8n-CAEMA detector with attention mechanisms for small-target detection.
- Implementation of a semi-supervised pipeline using Pseudo-Label Purification Filtering (PPLF) to reduce annotation dependence.
Main Results:
- The YOLOv8n-CAEMA detector achieved 0.840 mAP@0.5 with only 2.932 M parameters.
- GrainPest-SSL improved mAP@0.5 from 0.738 to 0.799 under a 30% labeled setting.
- The deployed detector achieved 13.6 FPS on a Jetson Orin Nano Dev Kit, demonstrating real-time performance.
Conclusions:
- GrainPest-SSL offers a balanced accuracy-efficiency solution for stored-grain pest detection under label-limited conditions.
- The framework supports smart granary systems with scheduled inspection, early screening, and intelligent warnings.
