Attention-guided Fine-grained Feature Learning For Robust Face Forgery Detection

Yizhe Zhu,Jialin Gao,Qiong Liu,Xi Zhou

DOI: https://doi.org/10.1109/ICPR56361.2022.9956159

2022-01-01

Abstract:With the swift development of deep learning, hyper-realistic images generated by advanced facial manipulation techniques have posed a serious threat to the trustworthiness of digital media information. Most existing approaches formulate face forgery detection as a coarse-grained classification problem. They still significantly focus on low-level semantic cues which are sensitive to common corruptions such as video compression and generalise poorly to unseen forgeries. To address this issue, we propose a novel fine-grained feature learning framework for face forgery detection. To be specific, we adopt fine-grained frequency decomposition via a patch-wise manner to extract more sufficient information hidden in the frequency domain. We also propose the depth-wise separable attention module to select more informative features and captures the fine-grained cues from different input spaces. Moreover, to further explore the essential discrepancies, attention-guided feature augment module is introduced to fully exploit the multi-domain relations and incorporate frequency features into spatial clues. Extensive experiments and visualizations on public datasets fully demonstrate the effectiveness and robustness of our method against the state-of-the-art competitors.