Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision

By Yaojie Liu, Amin Jourabloo, Xiaoming Liu. 2018

679 citations recorded by the source corpus.

Face anti-spoofing is crucial to prevent face recognition systems from a security breach. Previous deep learning approaches formulate face anti-spoofing as a binary classification problem. Many of them struggle to grasp adequate spoofing cues and generalize poorly. In this paper, we argue the importance of auxiliary supervision to guide the learning toward discriminative and generalizable cues. A CNN-RNN model is learned to estimate the face depth with pixel-wise supervision, and to estimate rPPG signals with sequence-wise supervision. The estimated depth and rPPG are fused to distinguish live vs. spoof faces. Further, we introduce a new face anti-spoofing database that covers a large range of illumination, subject, and pose variations. Experiments show that our model achieves the state-of-the-art results on both intra- and cross-database testing.

Subfields: Computer Vision and Pattern Recognition, Genetics, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Genetics

In Signal Processing

Citations in the corpus

Detection and Continual Learning of Novel Face Presentation Attacks (2021, background): “Similar to most sub-fields in computer vision, advances in deep learning, which is inspired by the nervous system [17], have led to significant face PAD improvements on benchmark datasets using convolutional neural network-based (CNN) end-to-end representation learning and classification [1, 7, 12, ”
Searching Central Difference Convolutional Networks for Face Anti-Spoofing (2020, methodology): “On the other hand, several video level CNN methods are presented to exploit the dynamic spatio-temporal [56, 62, 33] or rPPG [31, 36, 32] features for PAD.”
CASIA-SURF: A Large-Scale Multi-Modal Benchmark for Face Anti-Spoofing (2019, methodology): “In recent years, face PAD algorithms [26], [28] have achieved great performances.”
Face Anti-Spoofing: Model Matters, so Does Data (2019, methodology): “The performance of our model is comparable with the latest results of FaceDs [18], while slightly worse than those of Auxiliary [26].”
CelebA-Spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations (2020, methodology): “1% Auxiliary* [23] - 22.”
Face Anti-Spoofing with Human Material Perception (2020, methodology): “To distinguish live faces from spooﬁng faces, at the training stage, we normalize live depth maps and spooﬁng reﬂection maps in a range of [0 , 1], while setting spooﬁng depth maps and live reﬂection maps to 0, which is similar to [14,28].”
On the Detection of Digital Face Manipulation (2019, background): “i) Physical spoofing attacks can be as simple as face printed on a paper, replaying image/video on a phone, or as complicated as a 3D mask [8, 24, 34, 35].”
Learning One Class Representations for Face Presentation Attack Detection Using Multi-Channel Convolutional Neural Networks (2020, methodology): “Recently, CNN based methods have been more successful as compared to feature-based methods [16], [10], [17], [18].”
Multi-Adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection (2019, methodology): “We compare several state-of-the-art face anti-spoofing methods as follows: Multi-Scale LBP (MS LBP) [19] ; Binary CNN [32]; Image Distortion Analysis (IDA) [31]; Color Texture (CT) [5]; LBPTOP [23]; and Auxiliary [18]: This method learns a CNN-RNN model to estimate the face depth from one frame and ”
Deep Anomaly Detection for Generalized Face Anti-Spoofing (2019, methodology): “Finally, the third method is the one proposed in [23], which introduces a two-branch deep neural network that incorporates pixel-wise auxiliary supervision constrained by the depth reconstruction for all genuine samples (attacks are forced to belong to a plane) and the estimation of a remote…”
Multispectral Biometrics System Framework: Application to Presentation Attack Detection (2020, positive): “The suggested network architecture was inspired by the supervision channel approach in [70], [71] and its first part (identical to [67]) is equivalent to a patch-based architecture when the stride is 1, albeit with increased computational efficiency and reduced memory overhead.”
Face De-Spoofing: Anti-Spoofing via Noise Modeling (2018, methodology): “Table 4 shows the comparison of our method and the best 3 out of 18 previous methods [18,49].”
Single-Side Domain Generalization for Face Anti-Spoofing (2020, methodology): “61 Auxiliary [23] - 28.”
Deep Spatial Gradient and Temporal Depth Learning for Face Anti-Spoofing (2020, methodology): “Based on the different color changes between the living and spoofing face videos, rPPG [31, 34, 32] features are also explored for PAD.”