Deep Learning based Kinship Verification on KinFaceW-I Dataset
Hemprasad Yashwant Patil, Avinash Chandra · 2019
Face has been a prominent biometric trait and is widely appropriated for several tasks. One such a task is `kinship verification' which primarily focuses on examining whether a disposed image pair appertains to a same family or not Kinship verification system grabs a large apportion of applications which are advantageous to the society. A convolutional neural network specific algorithm which can function as classifier between `kin' and `non-kin' categories has been proposed in this work. The proposed approach incorporates a novel deep learning based layered neural network architecture. This approach is evident to enact superior performance than a few reported algorithms and the classification entirely depends upon CNN specific feature vectors. We have applied the proposed technique on the Kinship Face in Wild (KinFaceW) dataset version - I which can be publicly obtained. Under a standard experimentation protocol, we have achieved an average verification accuracy of 86.94%.