University of Economics and Business
 
The Improved Network Model Based on the YOLOv5 Method for Enhanced Safety Detection in Construction Sites

Ensuring the safety of workers is a crucial task in construction site management. Therefore, Dr. Tran Nguyen Ngoc Cuong, Prof. Dr. Tran Duc Hoc, Dr. Bui Dao Quang Thanh, and Nguyen Ngoc Thoan have proposed an improved detection network model based on the YOLOv5 method for safety warnings in construction sites in their study titled 'Improved detection network model based on YOLOv5 for warning safety in construction sites,' published in the International Journal of Construction Management, a Q1-ranked journal in the Web of Science.


Numerous accidents have occurred at construction sites due to various reasons such as slips, collisions, electric shocks, or entrapment in operating equipment. Therefore, the use of devices to ensure labor safety is highly necessary. Appropriate Personal Protective Equipment (PPE) is mentioned in widely utilized safety regulations to ensure worker safety. However, relying on traditional methods like physical monitoring and video observation for PPE usage can lead to time wastage, delayed response, and missed inspections. To overcome these limitations, the authors utilized a new algorithm called You Only Look Once (YOLO), specifically YOLOv5, which includes four network architectures: YOLOv5s, YOLOv5m, YOLOv5l, and YOLOv5x, for safety monitoring. The study employed a dataset of 11,978 samples to establish a digital safety monitoring system through training and testing phases. The comparative results among the four network architectures demonstrated that YOLOv5 performed the best with an average detection speed of 110 frames per second, meeting the real-time detection requirements.

The research has made valuable contributions to both theory and practice. Specifically:

1. It provides a solution for automatic recognition of Personal Protective Equipment (PPE) in construction sites.

2. It proposes a tool that serves as a valuable support for engineers in automated detection of PPE in construction sites.

3. The effectiveness and superiority of the method used in the study are highly regarded due to its implementation on a large dataset consisting of 11,978 samples and real-life construction scenarios.

The research helps expand the body of knowledge related to occupational safety management in the construction field. The findings of the study can assist managers in minimizing limitations in occupational safety management, thereby reducing workplace accidents in construction.

 

>> The details of the article can be found at:

 

AUTHOR INTRODUCTIONS

Dr. Tran Nguyen Ngoc Cuong - VNU University of Economics and Business, Vietnam National University, Hanoi

Assoc. Prof. Dr. Tran Duc Hoc - Ho Chi Minh City University of Technology, Vietnam National University, Ho Chi Minh City

Dr. Bui Dao Quang Thanh - Ho Chi Minh City University of Technology, Vietnam National University, Ho Chi Minh City

Nguyen Ngoc Thoan - Hanoi University of Civil Engineering


UEB Media