Manipulated Faces Detection with Adaptive Filter

With the progress of face manipulation techniques, synthesized faces are spreading on Internet, which raises concerns about potential threats. To prevent these techniques’ abuse, various detection algorithms have been proposed. In this paper, we consider image’s frequency information, then propose an adaptive filtering algorithm named spatial and (SAF) Network. SAF is a dual-stream network that considers domains. domain, wavelet transform used to divide image into different bands, filter introduced, aims capture more decisive information by giving weights frequencies. fuse features, pyramid pooling fusion (SPPF) proposed, solves mismatch feature maps, relationship between patches attention mechanism. Experiment results show performance better than comparison algorithm.