Learn Convolutional Neural Network for Face Anti-Spoofing

By Jianwei Yang, Zhen Lei, S. Li. arXiv.org, 2014

467 citations recorded by the source corpus.

Though having achieved some progresses, the hand-crafted texture features, e.g., LBP [23], LBP-TOP [11] are still unable to capture the most discriminative cues between genuine and fake faces. In this paper, instead of designing feature by ourselves, we rely on the deep convolutional neural network (CNN) to learn features of high discriminative ability in a supervised manner. Combined with some data pre-processing, the face anti-spoofing performance improves drastically. In the experiments, over 70% relative decrease of Half Total Error Rate (HTER) is achieved on two challenging datasets, CASIA [36] and REPLAY-ATTACK [7] compared with the state-of-the-art. Meanwhile, the experimental results from inter-tests between two datasets indicates CNN can obtain features with better generalization ability. Moreover, the nets trained using combined data from two datasets have less biases between two datasets.

Subfields: Aerospace Engineering, Information Systems, Signal Processing

View Paper: https://www.semanticscholar.org/paper/3c0b66ddf9f669bca73fe3c54f1287c514c42ce4

How researchers see this paper

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

In Aerospace Engineering

In Information Systems

In Signal Processing

Citations in the corpus

Face Spoofing Detection Using Colour Texture Analysis (2016, background): “However, these are shown to be important factors already in intra-database tests [6], [10], [56] and we noticed that they also significantly affect the cross-database performance.”
CASIA-SURF: A Large-Scale Multi-Modal Benchmark for Face Anti-Spoofing (2019, methodology): “State-of-the-art methods [27], [45], [55], [56] are listed for comparison, which use the ReplayAttack [8] dataset for training.”
Single-Side Domain Generalization for Face Anti-Spoofing (2020, methodology): “This is because all other face anti-spoofing methods [6, 9, 23, 26, 39, 42] except for the MADDG [33] method pay no attention to the intrinsic distribution relationship among different domains.”
Searching Central Difference Convolutional Networks for Face Anti-Spoofing (2020, methodology): “Most recent deep learning based FAS methods are usually built upon image classification task based backbones [61, 62, 20], such as VGG [54], ResNet [22] and DenseNet [23].”
Deep Learning in the Wild (2018, methodology): “In more recent work, deep learning methods have been introduced [3,64,63,31].”
Learning Deep Models for Face Anti-Spoofing: Binary or Auxiliary Supervision (2018, methodology): “50.2% 47.9% LBP [19] 55.9% 57.6% LBP-TOP [19] 49.7% 60.6% Motion-Mag [8] 50.1% 47.0% Spectral cubes [40] 34.4% 50.0% CNN [48] 48.5% 45.5% LBP [11] 47.0% 39.6% Colour Texture [12] 30.3% 37.7% Proposed method 27.6% 28.4% Furthermore, we utilize the CASIA-MFSD and Replay-Attack databases to perform…”
Deep Face Recognition: A Survey (2018, methodology): “[194] trained a CNN using both a single frame and multiple frames with five scales, and the live/spoof label is assigned as the output.”
Multi-Adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection (2019, methodology): “Li et al. [14] design a low-rank parameterized CNN model for end-to-end domain generalization learning.”
Deep Tree Learning for Zero-Shot Face Anti-Spoofing (2019, methodology): “Limitation of feature selection: They use handcrafted features such as LBP to represent live faces, which were shown to be less effective for known spoof detection [27, 32, 37, 48].”
Deep Models and Shortwave Infrared Information to Detect Face Presentation Attacks (2020, background): “As expected, there also exists approaches relying on deep Convolutional Neural Networks (CNN): relevant examples can be found in [13] and [14].”
Face De-Spoofing: Anti-Spoofing via Noise Modeling (2018, methodology): “Similar to texture-based methods, CNN-based methods [7,2,8] design a unified process of feature extraction and classification.”
Domain Generalization: A Survey (2021, background): “The WILDS Benchmark has been recently introduced, with a goal to study distribution shift faced in the wild [85].”
Capsule-forensics: Using Capsule Networks to Detect Forged Images and Videos (2018, methodology): “Chigovska et al. [7] 17.17 Pereira et al. [8] 08.51 Kim et al. [17] 12.50 Yang et al. [18] 02.30 Menotti et al. [19] 00.75 Alotabib et al. [20] 10.00 Ito et al. [21] 00.43 Nguyen et al. [12] 00.00 Capsule-Forensics 00.28 Capsule-Forensics-Noise 00.00”
Face Anti-Spoofing: Model Matters, so Does Data (2019, background): “The strong representation power of modern CNNs has been exploited [27] in face antispoofing research [16, 23, 31, 42].”
Cross-Domain Face Presentation Attack Detection via Multi-Domain Disentangled Representation Learning (2020, methodology): “…that leads to poor generalization ability of PAD models under new application scenarios. face PAs, which assume that there are inherent disparities between live and spoof faces, such as texture in color space [4, 6], image distortion [48], temporal variation [39] or deep semantic features [50, 36].”