Kinship Verification via Reference List Comparison

Kinship Verification via Reference List Comparison

연구 분야: Verification

학회: Chinese Conference on Biometric Recognition

Kinship verification based on facial images has attracted the attention of pattern recognition and computer vision community. Most of existing methods belong to supervised mode, in which they need to know the labels of training samples. In this paper, we adapt an unsupervised method via Reference List Comparison (RLC) for kinship verification task, which does not use external data or data augmentation. Specifically, we obtain a reference list by calculating the similarities of a probe image and all the images in the reference set. Given two probe face images, their similarity is reflected by the similarity of the two ordered reference lists. Experimental results on the KinFaceW-I and KinFaceW-II datasets show the effectiveness of RLC approach for kinship verification.