Image Inpainting Detection Based on High-Pass Filter Attention Network

Image inpainting based on deep learning has been greatly improved. The original purpose of image was to repair some broken photos, such as artifacts. However, it may also be used for malicious operations, destroying evidence. Therefore, detection and localization operations are essential. Recent research shows that high-pass filtering full convolutional network (HPFCN) is applied achieves good results. those methods did not consider the spatial location channel information feature map. To solve these shortcomings, we introduce squeezed excitation blocks (SE) propose a filter attention (HPACN). In extraction, apply concurrent (scSE) enhance extraction obtain more information. Channel (cSE) introduced in upsampling localization. experimental results show proposed method can achieve improvement ImageNet.