Image inpainting is an important research direction in the field of image processing.In order to solve the problem of fuzzy texture in the current image inpainting algorithm that the restoration area is inconsistent with the surrounding area,an image inpainting method is proposed.Based on the generative adversarial networks,this method introduces conditional features and self-attention modules,and associates the specific dimensions of data with semantic features.The inpainting model trained by this method can inpainting specific types of images and ensure the consistency of the overall inpainting and the rationality of local information details.Experimental results show that good results can be obtained by training and testing on CeleBA face datasets.