Face image occlusion repair algorithm based on deep learning and application research thereof

The invention discloses a face image occlusion restoration algorithm based on deep learning and an application research thereof in the field of face image occlusion restoration, which comprise research targets, research contents, key problems to be solved, research methods, technical routes, experimental schemes and feasibility analysis. The research content comprises a reference image guided shielded face repair model, repair model optimization based on a conditional generative adversarial network and an attention mechanism, and a learning algorithm of the repair model based on the conditional generative adversarial network and the attention mechanism; the key problems to be solved include introducing a reference guide image as a problem in a condition input constraint restoration process, selecting an appropriate convolution method to solve a large-scale irregular shielding problem, and a problem that a reference image attention layer is not enough to generate enough restoration information. The research me