Lightweight Helmet Detection Algorithm Using an Improved YOLOv4

Abstract

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.

Medical instrument detection with synthetically generated data

“…In the publication by Fujii, R. 6 , a neural network is also conditioned to recognise surgical instruments. Differs moderately from Ran, B. et al 1 and Lehr, R. et al 5 in the choice of background environment. The instruments to be identified are not located on an instrument tray, but in an open surgical procedure.…”

“…The other area of research in this field is the recognition of surgical instruments used in open surgery. Ran, B. et al 1 trained an artificial intelligence to recognise and identify 26 different instruments on an operating table. The system was used to support the staff in counting the instruments.…”

“…In order to reduce this workload, more and more working groups are researching ways to develop assistance systems, many of which utilize artificial intelligence methods. This also applies to the recognition of surgical instruments, with the aim of supporting, for example, counting tasks 1 or the sterilisation process. The data sets used, which must be as large and diverse as possible for robust recognition, have a decisive influence here.…”

Application development and technical analysis of rider helmet wearing monitoring system

“…A lighter network was used in place of the backbone network. PP-LCNet, and adopted depth-detachable convolution to reduce model parameters, which not only improved the accuracy and detection speed, but also greatly reduced the model size [11] . Pushkar used the YOLOv5 object detection model of transfer learning to build a rider helmet detection system.…”

Enhancing helmet and cigarette detection in electricity power construction based on Yolov5s-I algorithm

“…Researchers have proposed several more advanced algorithms for image recognition to detect target objects such as Gaussian Mixture Model (GMM) [2] , Histogram of Oriented Gradients (HOG) [3] and Support Vector Machines [4] . These algorithms are also widely used for helmet detection in construction scenarios to reduce the accident rate in the construction industry [5] . Skin color and Hu moments are used to identify helmets as in [6] , features of helmets of different colors are extracted and hierarchical support vector machines are used to classify workers wearing helmets as in [7] .…”