Uncertainty-aware hierarchical labeling for face forgery detection

Bingyao Yu, Wanhua Li, Li Xiu, Jie Zhou, Jiwen Lu

Abstract: Highlights•We are the first to introduce uncertainty learning for face forgery detection.•We propose the hierarchical labeling as auxiliary supervised signals to exploit the multi-level information to maximize the generalization of our framework.•Experimental results on several datasets show that our proposed approach achieves very competitive performance and strong generalization ability.