Related Experiment Videos
Violation detection in power operation sites based on multi-scale detection and few-shot learning
Yaokuan Wen1, Jun Wang1, Qiming Liu1
1State Grid Henan Marketing Service Center, Zhengzhou, China.
Frontiers in Artificial Intelligence
|July 1, 2026
Summary
This study introduces a novel framework for detecting safety violations at power operation sites, improving small-object detection and adapting to limited data. The new method enhances worker safety and reliable electricity supply through intelligent monitoring.
Area of Science:
- Electrical Engineering
- Computer Science
- Artificial Intelligence
Background:
- Safety supervision is critical in power operation sites for worker safety and reliable electricity.
- Current safety violation detection methods struggle with limited labeled data, small object detection, and complex backgrounds.
Purpose of the Study:
- To develop an advanced framework for safety violation detection in power operation sites.
- To address limitations of existing methods, including insufficient labeled data and poor small-object detection.
Main Methods:
- Integration of multi-scale object detection with few-shot learning.
- Utilized a feature pyramid network and channel attention for enhanced small object perception.
- Employed a meta-learning strategy within a few-shot learning framework to handle data scarcity.
Main Results:
- The proposed method significantly outperforms existing approaches across various metrics.
- Demonstrated notable improvements in small-object detection accuracy and few-shot learning performance.
- Achieved enhanced detection accuracy, robustness, and generalization capabilities.
Conclusions:
- The framework effectively tackles safety violation detection challenges in power operations.
- Offers a practical solution for intelligent safety monitoring with high potential for real-world deployment.
