Aiming at the problem of low detection accuracy due to the difficulty of small target detection in the mask wearing detection task in public places, this paper proposes a mask wearing detection algorithm based on improved YOLOv4. Firstly, based on the original 3-scale feature map of YOLOv4, the feature map obtained by four times downsampling is introduced to add small target semantic information to the feature fusion module. Secondly, combined with the advantages of DenseNet to enhance the transmission of features, the feature fusion module is improved by using the idea of dense connection, so that each output feature layer can be fully integrated into the semantic information of small targets. The comparison results show that the average accuracy of the improved algorithm is as high as 94.77%, which is 1.25% higher than that of the original YOLOv4 network. Compared with other mainstream algorithms, it also has better detection results in mask wearing detection.
Research on application of mask wearing based on YOLOv4 model
12.10.2022
1073793 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Application of lightweight YOLOv4 in vehicle detection
British Library Conference Proceedings | 2022
|Pedestrian Detection Algorithm Based on Improved YOLOv4
IEEE | 2023
|