Generalized Face Anti-Spoofing via Cross-Adversarial Disentanglement with Mixing Augmentation

Huang, Hanye; Xiang, Youjun; Yang, Guodong; Lv, Lingling; Li, Xianfeng; Weng, Zichun; Fu, Yuli

doi:10.1109/icassp43922.2022.9746716

Cited by 10 publications

(4 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A two-stream network is utilized to fuse the input RGB image and meta-pattern learning was proposed to improve the generalization [13]. A cross-adversarial training scheme is proposed to improve the generalization by minimizing the correlation among two sets of features [16]. The work reported in [14], aims to learn a generalized feature space by designing the target data to the source-domain style and called Generative Domain Adaptation (GDA).…”

Section: Literature Reviewmentioning

confidence: 99%

“…To improve the generalization, the majority of recent approaches in face PAD such as adversarial learning [12], meta pattern learning [13], generative domain adaptation [14], hypothesis verification [15], or cross-adversarial learning [16], address the domain generalization issue by exploiting a common feature space from multiple source domains, but the performance remains limited due to a substantial distribution difference among source domains. For instance, research in [17] relies on a shared feature space and assumes that it would also be invariant to domain shift.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection

Usman¹,

Beddiar²,

Oussalah³

2023

Preprint

View full text Add to dashboard Cite

Face presentation attack detection (PAD) plays a pivotal role in securing face recognition systems against spoofing attacks. Although great progress has been made in designing face PAD methods, developing a model that can generalize well to an unseen test domain remains a significant challenge. Moreover, due to different types of spoofing attacks, creating a dataset with a sufficient number of samples for training deep neural networks is a laborious task. This work addresses these challenges by creating synthetic data and introducing a deep learning-based unified framework for improving the generalization ability of the face PAD. In particular, synthetic data is generated by proposing a video distillation technique that blends a spatiotemporal warped image with a still image based on alpha compositing. Since the proposed synthetic samples can be generated by increasing different alpha weights, we train multiple classifiers by taking the advantage of a specific type of ensemble learning known as a stacked ensemble, where each such classifier becomes an expert in its own domain but a non-expert to others. Motivated by this, a meta-classifier is employed to learn from these experts collaboratively so that when developing an ensemble, they can leverage complementary information from each other to better tackle or be more useful for an unseen target domain. Experimental results using half total error rates (HTERs) on four PAD databases CASIA-MFSD (6.97%), Replay-Attack (33.49%), MSU-MFSD (4.02%), and OULU-NPU (10.91%)) demonstrate the robustness of the method and open up new possibilities for advancing presentation attack detection using ensemble learning with large-scale synthetic data.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection

Usman¹,

Beddiar²,

Oussalah³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The results confirm the effectiveness of our approach. In Table 3, we compare our results with those state-of-the-art methods [25][26][27][28][29][30][31] which focus on domain generalization settings. Based on the experimental results, our method provides best performance on two benchmark databases (M and C) and a competitive performance on third dataset (I).…”

Section: Implementation Detailsmentioning

confidence: 99%

“…Moreover, state-of-the-art approaches address the domain generalization issue by aligning the feature distribution between source and target domain. For instance, adversarial learning [25], meta pattern or meta-teacher learning [8,26], cross-adversarial learning [27], generative domain adaptation [28], hypothesis verification [29], or shuffled style assembly network [30] were introduced for improving the generalization ability of face anti-spoofing. However, most of them typically fit the models to a shared feature space which remains always challenging for a better generalization.…”

mentioning

confidence: 99%

Face anti‐spoofing from the perspective of data sampling

Muhammad

Oussalah

2022

Electronics Letters

View full text Add to dashboard Cite

Without deploying face anti-spoofing countermeasures, face recognition systems can be spoofed by presenting a printed photo, a video, or a silicon mask of a genuine user. Thus, face presentation attack detection (PAD) plays a vital role in providing secure facial access to digital devices. Most existing video-based PAD countermeasures lack the ability to cope with long-range temporal variations in videos. Moreover, the key-frame sampling prior to the feature extraction step has not been widely studied in the face anti-spoofing domain. To mitigate these issues, this paper provides a data sampling approach by proposing a video processing scheme that models the long-range temporal variations based on Gaussian weighting function (GWF). Specifically, the proposed scheme encodes the consecutive t frames of video sequences into a single RGB image based on a Gaussian-weighted summation of the t frames. Using simply the data sampling scheme alone, it is demonstrated here that state-of-the-art performance can be achieved without any bells and whistles in both intra-database and inter-database testing scenarios for the three public benchmark datasets; namely, replay-Attack, MSU-MFSD, and CASIA-FASD. In particular, the proposed scheme provides a much lower error (from 15.2% to 7.6% on CASIA-FASD and 5.9% to 4.9% on replay-attack) compared to baselines in cross-database scenarios.

show abstract

One-class learning for face anti-spoofing via pseudo-negative sampling

Gwon,

Kim

2023

Multimed Tools Appl

View full text Add to dashboard Cite

Generalized Face Anti-Spoofing via Cross-Adversarial Disentanglement with Mixing Augmentation

Cited by 10 publications

References 14 publications

Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection

Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection

Face anti‐spoofing from the perspective of data sampling

One-class learning for face anti-spoofing via pseudo-negative sampling

Contact Info

Product

Resources

About