期刊名称：International Journal of Hybrid Information Technology
印刷版ISSN：1738-9968
出版年度：2015
卷号：8
期号：11
页码：285-296
DOI：10.14257/ijhit.2015.8.11.24
出版社：SERSC
摘要：To achieve the problem of partial occlusion and multi-pose in the face detection, a face detection algorithm based on deep learning is proposed. Through establishing deep model, the probabilistic correlations of the visibilities are learned in different face local region. Firstly, face local regions are detected by part-based face detector. Secondly, the detection results are feed to deep model, which describes the correlations of different regions. Finally, the face detection is completed according to the correlations. The experimental results show that the proposed algorithm is effective in the case of occlusion and multi-pose.
关键词：Face detection; deep learning; deep model; part-based; detection rate; false ; positive rate; recall rate
