基于卷积神经网络级联的蒙面人脸检测算法

Masked Face Detection Algorithm Based on CNN Cascade

精确地检测蒙面人脸是鉴别和追踪罪犯或者恐怖分子的重要手段,因此,一个高效的蒙面人脸检测算法这对于打击犯罪,维护社会治安稳定有重要意义。然而,由于面具遮挡导致的人脸信息缺失,传统的人脸检测算法很难取得令人满意的结果。针对这一问题,在本论文中,我们提出了一种适用于蒙面人脸检测的卷积神经网络级联算法。该级联网络共有三级,在训练时,第一级采用整个训练样本集进行训练,之后逐级对前一级训练中分类错误的样本进行训练,以获得更强的辨别能力。这一策略也能避免第一级网络的过度拟合。为了进一步保证算法的检测精度,我们采用迁移学习的方法,利用大型的通用人脸数据集和蒙面人脸数据集来训练和微调分类网络模型。此外,我们还优化了每一级的网络结构,从而提高了计算效率。我们在蒙面人脸的测试数据集上对算法进行测试。实验结果表明,我们在87. 8%的召回率下取得了86. 6%的精确率。并且,相比于传统的卷积神经网络算法,我们的方法具有较高的检测精度和检测效率。

Accurately detecting masked faces can be applied on tracking and identifying criminals or terrorists. As a unique face detection task, masked face detection is much more difficult because of extreme occlusions which leads to the loss of face details. In this paper, we propose a new CNN based cascade framework. It consists three stages of CNN. We use all of the training data to train the first CNN, then use the wrongly classified data to train the later stage CNNs to achieve greater classification ability. To ensure the accuracy of this algorithm we apply the migration leanfing method, we propose a new masked face dataset to fine -tune the classification networks. Besides we opti- mize the structure of the classification networks to improve the detection efficiency. Experimental results show that our masked face detection algorithm has great pefformance, we achieve the precision of 86.6% at 87.8 recall. This has great significance in combating crime and maintaining public security.