耳朵信息对侧脸检测的影响研究

The Influence of Ear Biometric on Profile Face Detection

如何准确地检测和定位图像中的人脸是人脸检测领域的关键问题.为了进一步提高人脸检测器的性能,常见的方法是增加训练数据集或采用更鲁棒的人脸特征表示,而训练人脸检测器的一个基础工作是:为训练图像中的人脸标注边界框.但标注的人脸边界框是否应该包含耳朵信息,以及对训练出的侧脸检测器性能的影响尚未被研究.本文的实验结果表明:在侧脸数据集上训练人脸检测器时,如果人脸边界框包含耳朵信息,基于DPM(Deformable Parts Model)方法训练得到的侧脸检测器使侧脸检测的准确率降低1.9%,召回率提高6.3%.而基于Viola&Jones和Fast R-CNN方法训练得到的侧脸检测器使准确率分别提高6.8%和4.4%,召回率分别提高14.9%和12.9%.这说明包含耳朵信息训练出的侧脸检测模型,有助于提高侧脸检测率.

In the field of face detection,how to accurately detect and locate faces in the images is a key problem.In order to improve the performance of face detector,the common approach is to add more training datasets or use more robust facial feature representations.Yet,one of the foundational tasks in training face detector is the annotation of the face bounding boxes for the training images.Should the face bounding box include ear information?What is the corresponding influence on the performance of profile face detection?This is an uninvestigated issue.Our experimental results show that,when using profile face datasets and the faces bounding include ear to train the face detector,the accuracy of profile face detector based on DPM(the Deformable Parts Model)decreases by 1.9%,yet,the recall increases by 6.3%.In the case of Viola&Jones and Fast R-CNN,the improvements are significant,the corresponding accuracy increases by 6.8%and 4.4%,and recall increases by 14.9%and 12.9%,respectively.This reveals that,when training profile face detection model with the ear information,it can significantly improve the recall of profile face detection.