A Multiscale Grouped Convolution and Lightweight Adaptive Downsampling-Based Detection of Protective Equipment for Power Workers

Xin Liu,Yingna Li

DOI: https://doi.org/10.3390/electronics13112079

IF: 2.9

2024-05-28

Electronics

Abstract:Convolutional neural network-based detection models have been extensively applied in industrial production for monitoring the use of safety protection equipment, ensuring worker safety. This paper addresses scenarios in electrical operations where the safety protection requirements are more comprehensive and stringent. This paper proposes an improved detection model, ML-YOLOv8n-Light, based on YOLOv8n, targeting issues of low detection efficiency and large model size that make deployment challenging in current safety protection equipment-wearing detection models for electrical operations. Our model confronts the disparity in safety protection equipment sizes with a novel, lightweight multi-scale grouped convolution (MSGC) scheme integrated into the architecture. A lightweight adaptive weight downsampling (LAWD) mechanism is also fashioned to replace the traditional downsampling methods, optimizing resource consumption without sacrificing performance. Additionally, to enhance the detection fidelity of smaller items, such as insulated gloves, we added feature-rich shallow maps and a dedicated detection head for such objects. To enhance the detection efficiency of YOLOv8n, inspired by part convolution, we improved the spatial pyramid pooling fast (SPPF) and the detection heads. The experiments conducted on the custom dataset power safe attire dataset (PSAD) showed that compared to the original model, mAP50 increased by 2.1%, mAP50-95 by 3.1%, with a 29% reduction in parameters, an 18% reduction in computations, and a 23% compression of the model size. There are fewer detection omissions at long distances and under occlusion, fewer false positives, and computing resources are allocated more efficiently.

engineering, electrical & electronic,physics, applied,computer science, information systems

What problem does this paper attempt to address?

### Problems Addressed by the Paper

This paper aims to address the issues of inefficient detection and large model size in the detection of protective equipment for power workers. Specifically:

1. **Existing Problems**:
- The current models used for detecting the wearing of protective equipment in power operation scenarios suffer from low detection efficiency, large model size, and deployment difficulties.
- In the power industry, the requirements for protective equipment are more comprehensive and stringent, and existing detection models struggle to meet these demands.

2. **Proposed Solutions**:
- The paper proposes an improved detection model, ML-YOLOv8n-Light, which is optimized based on YOLOv8n.
- A new lightweight multi-scale grouped convolution (MSGC) scheme is introduced to handle the differences in protective equipment of various sizes.
- A lightweight adaptive weight downsampling (LAWD) mechanism is designed to replace traditional downsampling methods, optimizing resource consumption without sacrificing performance.
- For the detection of small items (such as insulating gloves), feature-rich shallow maps and dedicated detection heads are added.

3. **Experimental Results**:
- Experiments on the custom dataset PSAD show that compared to the original model, mAP50 increased by 2.1%, mAP50-95 increased by 3.1%, parameters were reduced by 29%, computation was reduced by 18%, and model size was compressed by 23%.
- The phenomenon of missed detections in long-distance and occlusion scenarios was significantly reduced, false positives were also reduced, and the allocation of computational resources became more efficient.