Suppressing Spoof-Irrelevant Factors for Domain-Agnostic Face Anti-Spoofing

Face anti-spoofing aims to prevent false authentications of face recognition systems by distinguishing whether an image is originated from a human or spoof medium. In this work, we note that images unseen domains having different spoof-irrelevant factors (e.g., background patterns and subject) induce domain shift between source target distributions. Also, when the same SiFs are shared genuine images, they show higher level visual similarity hinders accurate anti-spoofing. Hence, aim minimize discrepancies among via alleviating effects SiFs, achieve improvements in generalization domains. To realize our goal, propose novel method called Doubly Adversarial Suppression Network (DASN) trained neglect irrelevant focus more on faithful task-relevant factors. Our DASN consists two types adversarial learning schemes. first scheme, multiple suppressed deploying discrimination heads against encoder. second each also adversarially suppress factor, group secondary classifier encoder intensify factor overcoming suppression. We evaluate proposed four public benchmark datasets, remarkable evaluation results generalizing The demonstrate effectiveness method.