Mask wearing detection based on YOLOv3
Xiaokang Ren, Xingxing Liu · Journal of Physics Conference Series · 2020
Abstract The CoVID-19 is still raging around the world, and the work of relying on manpower to detect masks in public places is time-consuming and laborious. To solve this problem, this paper proposes an improved Face_mask Net detection method for convolutional neural network based on YOLOv3. First, a new convolutional neural network Face_mask Net is designed based on YOLOv3 network structure. Second, before training, Face_mask Net uses the K-Means algorithm to cluster the labeled dataset and change its anchor value. Finally, the Face_mask Net loss function uses DIoU and the classifier uses DIoU-NMS. The combination of the two can further improve the detection accuracy of the target. To verify the effectiveness of the proposed algorithm, the Face_mask Dataset for mask detection was collected and annotated in this paper. The experimental results show that Face_mask Net can effectively detect the target of wearing mask and not wearing mask, and its performance can be close to real-time and the accuracy is higher than that of the network before the improvement.