Safety Helmet Wearing Detection Based on A Lightweight YOLOv4 Algorithm

Chen, Junhua; Deng, Sihao; Huang, Xueda; Yang, Xinrui; Dong, Yan

doi:10.1109/ithings-greencom-cpscom-smartdata-cybermatics55523.2022.00094

Cited by 1 publication

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the FPS of the improved YOLOv4 is not as good as that of YOLOv4-Tiny [ 42 ] and its model size is larger than that of YOLOv4-Tiny, its P, R, F1, and mAP are 6.66%, 9.25%, 8.5%, and 11.72% higher than those of YOLOv4-Tiny, respectively. Compared with Ghost-YOLOv4 [ 43 ], which replaces the YOLOv4 backbone with GhostNet, the improved YOLOv4 has 2.02%, 2.28%, 3.15%, and 6.64 pictures/s increments in the P, R, F1, mAP, and FPS, respectively, and a 1.72M decrement in model size. For faster R-CNN, CenterNet, SSD, and EffificientDet-D2 [ 29 ], the improved YOLOv4 is superior to them with regard to detection accuracy, speed, and model size.…”

Section: Experiments and Analysismentioning

confidence: 99%

Lightweight Helmet Detection Algorithm Using an Improved YOLOv4

Chen

Deng

Wang

et al. 2023

Sensors

Self Cite

View full text Add to dashboard Cite

Safety helmet wearing plays a major role in protecting the safety of workers in industry and construction, so a real-time helmet wearing detection technology is very necessary. This paper proposes an improved YOLOv4 algorithm to achieve real-time and efficient safety helmet wearing detection. The improved YOLOv4 algorithm adopts a lightweight network PP-LCNet as the backbone network and uses deepwise separable convolution to decrease the model parameters. Besides, the coordinate attention mechanism module is embedded in the three output feature layers of the backbone network to enhance the feature information, and an improved feature fusion structure is designed to fuse the target information. In terms of the loss function, we use a new SIoU loss function that fuses directional information to increase detection precision. The experimental findings demonstrate that the improved YOLOv4 algorithm achieves an accuracy of 92.98%, a model size of 41.88 M, and a detection speed of 43.23 pictures/s. Compared with the original YOLOv4, the accuracy increases by 0.52%, the model size decreases by about 83%, and the detection speed increases by 88%. Compared with other existing methods, it performs better in terms of precision and speed.

show abstract

Section: Experiments and Analysismentioning

confidence: 99%