摘要
很多工厂要求员工在进入工作空间之前佩戴口罩,而现有的识别模型大都存在扰动环境下准确率低、健壮性差的问题。针对以上问题,设计一种人脸口罩识别模型——AdaMask,以VGGNet-16模型为基础,改进模型训练机制和激活函数的参数,并且自行采集数据对其进行训练,使模型达到较高的识别准确率。通过对比实验证明,相对于其他模型,AdaMask有较高的健壮性。
Many factories require employees to wear masks before entering the work space,and most of the existing identification models have problems with low accuracy and poor robustness in the disturbance environment.In response to the above problems,a human face mask recognition model is designed-AdaMask,based on the VGGNet-16 model,improves the parameters of the model training mechanism and activation function,and we collect data to train it to make the model reach higher Accuracy of identification.Compared with other models,AdaMask has a high robustness compared to other models.
作者
刘文瑶
LIU Wenyao(Dongguan City University,Dongguan Guangdong 523419,China)
出处
《信息与电脑》
2023年第15期147-150,共4页
Information & Computer
关键词
人工智能模型
口罩识别
健壮性
artificial intelligence model
Mask recognition
robustness
