Kinship Recognition based on Deep Scattering Wavelet Convolutional Neural Network on Wild Facial Image

Kingship verification is a process that two or more people has a family relation such as father and son or other family relation. Numerous studies have been presented to investigate the relationship between people. Kingship verification can be done based on image of face. Most of the methods presented on face images work well on face data sets recorded under controlled conditions. However, due to the complex nature of environments, rapidly and accurately examining human kinship in real-world unrestricted or wild-type scenarios is still a challenging research. In this paper, in order to overcome the aforementioned challenges, an efficient and new method is presented. In the proposed method, a method is used to launch the operation to create a map. The created feature map is stable against deformation, transition, scaling, direction and Dilation in wild images. Group-Face and TSKinFace databases are used for simulation. In order to evaluate the evaluation of the proposed method, average recall of ۹۴.۱, precision ۹۴.۶, accuracy ۹۴.۷, specificity ۹۳.۸, and finally F_Measure ۹۵.۰ were used. The superiority of the proposed method in all comparisons shows the effectiveness of the proposed method in diagnosing kinship.

Keywords:

kinship , deep learning , Dispersion wavelet , Convolution Neural Network