A Performance Evaluation of Convolutional Neural Networks for Face Anti Spoofing

Abstract

In the current era, biometric based access control is becoming more popular due to its simplicity and ease to use by the users. It reduces the manual work of identity recognition and facilitates the automatic processing. The face is one of the most important biometric visual information that can be easily captured without user cooperation in an uncontrolled environment. Precise detection of spoofed faces should be on the high priority to make face based identity recognition and access control robust against possible attacks. The recently evolved Convolutional Neural Network (CNN) based deep learning technique has proven as one of the excellent method to deal with the visual information very effectively. The CNN learns the hierarchical features at intermediate layers automatically from the data. Several CNN based methods such as Inception and ResNet have shown outstanding performance for image classification problem. This paper does a performance evaluation of CNNs for face antispoofing. The Inception and ResNet CNN architectures are used in this study. The results are computed over benchmark MSU Mobile Face Spoofing Database. The experiments are done by considering the different aspects such as the depth of the model, random weight initialization vs weight transfer, fine tuning vs training from scratch and different learning rate. The favorable results are obtained using these CNN architectures for face antispoofing in different settings.

A Face Spoofing Detection Method Based on Domain Adaptation and Lossless Size Adaptation

“…Another useful clue in the recaptured faces is the degradation of color and image quality, which can be represented by the color texture histogram features [14], the color moments feature [25], and image quality assessment features [16], [17]. Recently, the CNNs [4]- [6] and the FCNs [7]- [10] were used for directly learning the classifier as well as the features for face spoofing detection task. These deep learning-based methods will be reviewed in Section II-C-II-D in detail.…”

“…Rehman et al [5] trained an 11-layer VGG network with its two derivations for face anti-spoofing in an end-toend scheme. Nagpal et al [6] explored deeper ResNet and GoogLeNet for training the face spoof detector. In the above work, decision of the CNNs are based on the whole face crops.…”

“…Recently, deep learning-based face spoofing detection has become popular. For example, some Convolutional Neural Networks (CNNs) [4]- [6] and some Fully Convolutional Networks (FCNs) [7]- [10] are used for process static frames. Some Recurrent Neural networks (RNNs) [7] are used for aggregating all the frames in the video.…”

Dual CNN Models for Unsupervised Monocular Depth Estimation

“…The depth map from images can be estimated in various ways like structure from motion [14], multi-view stereo [19], monocular methods [17], singleimage methods [18], etc. The deep learning and convolutional neural networks (CNNs) based methods perform outstanding in most of the problems of computer vision such as image classification [10], facial micro-expression recognition [15], face anti-spoofing [13], hyper-spectral image classification [16], image-to-image transformation [9], colon cancer nuclei classification [1], etc. Inspired from the success of deep learning, several researchers also tried to utilize the CNN for the depth prediction, specially in monocular imaging conditions.…”

PSNet: Parametric Sigmoid Norm Based CNN for Face Recognition

“…It also won the ImageNet Large Scale Challenge in 2012. After that, various CNN based architectures have been introduced for different applications such as VGGNet [2], GoogleNet [3], ResNet [4] for image classification; Local Bit-plane Decoded CNN [5], ChexNet [6] and RCCNet [7] for biomedial image retrieval/classification; CNN for face spoof detection [8], [9]; and many more.…”