Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Automatic Object Detection of Construction Workers and Machinery Based on Improved YOLOv5
https://orcid.org/http://orcid.org/0000-0001-5381-1905
https://orcid.org/http://orcid.org/0000-0003-4563-8860
https://orcid.org/http://orcid.org/0000-0003-1600-0155
https://orcid.org/http://orcid.org/0000-0001-9357-9522
https://orcid.org/http://orcid.org/0000-0002-0071-3137
Automatic detection and localization of workers and machinery on construction sites through surveillance video is important to supervise on-site safety and construction process, which could develop civil construction management and services. However, it is difficult to detect all instances due to the extremely complex construction environment and numerous multi-scale objects. This paper proposes an improved YOLOv5 model to automatically detect and localize construction workers and 11 common types of construction machinery. Firstly, use the bidirectional feature pyramid network (BiFPN) layer for better multi-scale feature fusion ability; Secondly, 3 × 3 convolution layer is replaced by RepVGG block, which shows favorable accuracy-speed trade-off. The experimental results indicate that the mAP (mean Average Precision) of our proposed method is 87.32%, which is 2.12% higher, and inference time reduce to 5.7 ms per frame.
Automatic Object Detection of Construction Workers and Machinery Based on Improved YOLOv5
https://orcid.org/http://orcid.org/0000-0001-5381-1905
https://orcid.org/http://orcid.org/0000-0003-4563-8860
https://orcid.org/http://orcid.org/0000-0003-1600-0155
https://orcid.org/http://orcid.org/0000-0001-9357-9522
https://orcid.org/http://orcid.org/0000-0002-0071-3137
Automatic detection and localization of workers and machinery on construction sites through surveillance video is important to supervise on-site safety and construction process, which could develop civil construction management and services. However, it is difficult to detect all instances due to the extremely complex construction environment and numerous multi-scale objects. This paper proposes an improved YOLOv5 model to automatically detect and localize construction workers and 11 common types of construction machinery. Firstly, use the bidirectional feature pyramid network (BiFPN) layer for better multi-scale feature fusion ability; Secondly, 3 × 3 convolution layer is replaced by RepVGG block, which shows favorable accuracy-speed trade-off. The experimental results indicate that the mAP (mean Average Precision) of our proposed method is 87.32%, which is 2.12% higher, and inference time reduce to 5.7 ms per frame.
Automatic Object Detection of Construction Workers and Machinery Based on Improved YOLOv5
https://orcid.org/http://orcid.org/0000-0001-5381-1905
https://orcid.org/http://orcid.org/0000-0003-4563-8860
https://orcid.org/http://orcid.org/0000-0003-1600-0155
https://orcid.org/http://orcid.org/0000-0001-9357-9522
https://orcid.org/http://orcid.org/0000-0002-0071-3137
Automatic detection and localization of workers and machinery on construction sites through surveillance video is important to supervise on-site safety and construction process, which could develop civil construction management and services. However, it is difficult to detect all instances due to the extremely complex construction environment and numerous multi-scale objects. This paper proposes an improved YOLOv5 model to automatically detect and localize construction workers and 11 common types of construction machinery. Firstly, use the bidirectional feature pyramid network (BiFPN) layer for better multi-scale feature fusion ability; Secondly, 3 × 3 convolution layer is replaced by RepVGG block, which shows favorable accuracy-speed trade-off. The experimental results indicate that the mAP (mean Average Precision) of our proposed method is 87.32%, which is 2.12% higher, and inference time reduce to 5.7 ms per frame.
Automatic Object Detection of Construction Workers and Machinery Based on Improved YOLOv5
Lecture Notes in Civil Engineering
Guo, Wei (Herausgeber:in) / Qian, Kai (Herausgeber:in) / Xiang, Yuanzhi (Autor:in) / Zhao, Jiayue (Autor:in) / Wu, Wenjing (Autor:in) / Wen, Caifeng (Autor:in) / Cao, Yunzhong (Autor:in)
International Conference on Green Building, Civil Engineering and Smart City ; 2022 ; Guilin, China
08.09.2022
9 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Object Detection for Construction Waste Based on an Improved YOLOv5 Model
| DOAJ | 2022
Improved detection network model based on YOLOv5 for warning safety in construction sites
| Taylor & Francis Verlag | 2024