基于改进SSD的人脸口罩佩戴检测算法

Face Mask Detection Algorithm based on Improved SSD

本文提出一种多尺度卷积特征融合目标检测方法,用于优化SSD模型对口罩遮挡、尺度变化、样式多样化以及小目标问题的检测精度。基准网络选用表征能力更强的残差网络,引入跳跃连接机制降低提取特征的冗余度,解决层数增加出现性能退化问题;同时引入低层边缘信息与高层语义信息的多尺度特征融合机制充分利用特征细节信息,最终通过参数再训练方式获得改进的SSD模型。实验结果表明,该方法在人脸口罩数据集测试精度90.65%,与原SSD算法82.37%提高8.28%,与SSD使用ResNet-50的87.99%提高2.66%。

This paper proposes a multi-scale convolution feature fusion target detection method to optimize the detection accuracy of the SSD model for mask occlusion,scale change,style diversification,and small target problems.The benchmark network selects a residual network with stronger characterization ability,introduces a jump connection mechanism to reduce the redundancy of extracted features,and solves the problem of performance degradation due to an increase in the number of layers;at the same time,it introduces a multi-scale feature fusion mechanism of low-level edge information and high-level semantic information to make full use of The characteristic detail information is finally obtained through parameter retraining to obtain an improved SSD model.Experimental results show that the test accuracy of this method on the face mask data set is 90.65%,which is 8.28%higher than the original SSD algorithm 82.37%,and 87.99%using ResNet-50 with SSD is 2.66%higher.