Facial image inpainting for big data using an effective attention mechanism and a convolutional neural network.

Big data facial image is an important identity information for people. However, facial image inpainting using existing deep learning methods has some problems such as insufficient feature mining and incomplete semantic expression, leading to output image artifacts or fuzzy textures. Therefore, it is of practical significance to study how to effectively restore an incomplete facial image.