Generating Realistic and Coherent Textures for Missing Regions in Images using DeepFill

Abstract: DeepFill is a method that allows you to paint an image with a free-form mask using a generative adversarial network (GAN). This includes contextual awareness and gated convolution to generate realistic and consistent textures for missing regions. It can handle various remediation scenarios such as: B. Delete objects, complete faces, delete text, etc. User-guided repairs can also be supported with additional inputs such as sketches and colors. DeepFill is based on his Jiahui Yu et al article "Free-Form Image Inpainting with Gated Convolution" published at ICCV 2019. Image inpainting uses information from surrounding pixels to fill in missing or damaged areas of the image. Image restoration can be used for many purposes. B. Restore corrupted images, remove unwanted objects, create artistic effects, etc.

Keywords: Inpainting, DeepFill, Image Processing.