Research And Application Of Image Inpainting Technology Based On Generative Confrontation Network

Deep learning is an extremely important part of the field of artificial intelligence and has been applied to hundreds of practical problems in recent years.The traditional image restoration technology has many defects in the image restoration effect.Compared with the traditional image restoration technology,these problems are largely solved after combining deep learning technology with image restoration technology.For example,OpenCV,which is widely used in image restoration technology,has obvious improvement in many aspects such as repairing effects compared with traditional image restoration technology.With the birth of the Generative Adversarial Nets,with the powerful capabilities of its generators,it brings more possibilities to the field of image restoration.The technical principle of generating adversarial network is to gradually improve the training of the generator model in the constant adversarial between the generator and the discriminator.Finally,the generator can generate a sample of the fake image.The main research contents of this paper include:1)Analysis and research on the original Generative Adversarial Nets.2)According to the Wasserstein distance idea,the WGAN network model is introduced to analyze the improvement of the original model against the network.3)Try to apply the generated anti-network technology to the image restoration technology and optimize the original image restoration technology.The main work of this paper includes:1)detailed analysis and research on the problems existing in the original generation of anti-network;2)attempt to apply the WGAN network model in image restoration technology;3)in the image restoration technology,combined with context information and the concept of perception and Poisson fusion ideas to optimize the image restoration technology,Make the repaired image perform better in terms of semantic continuity.The final experiment shows that the improved image restoration technology has more significant repair effects in all aspects than the traditional image restoration algorithm.