No abstract

Kinship verification from facial images is an interesting and challenging problem. The current algorithms on this topic typically represent faces with multiple low-level features, followed by a shallow learning model. In this paper, we propose to extract high-level features for kinship verification based on deep convolutional neural networks. Our method is end-toend, without complex pre-processing often used in traditional methods. The high-level features are produced from the neuron activations of the last hidden layer, and then fed into a soft-max classifier to verify the kinship of two persons.We first propose a basic structure of CNN (CNN-Basic) contains three convolutional layers, followed by a fully-connected layer and a soft-max layer. As shown in Figure 1, the input is a pair of 64 × 64 images with three channels (RGB). Following the input, the first convolutional layer is generated after convolving the input via 16 filters with a stride of 1. Each filter is with the size 5 × 5 × 6. The second convolutional layer filters the input of the previous layer with 64 kernels of size 5 × 5 × 16. The third convolutional layer contains 128 kernels of the size 5 × 5 × 64. After the convolutional layers, a fully-connected layer projects the extracted features into a subspace with 640 neurons. Max-pooling layers follow the first and second convolutional layers. Finally, this network is trained via a two-way soft-max classifier at the top layer.We adopt the ReLU function [1] as the activation function of the convolution layers, which has been shown to achieve better performance than the sigmoid function. With ReLU, the convolution operation is formulated as y j(r) = max 0, b j(r) + ∑

No abstract

No abstract

No abstract