Image Inpainting (a.k.a. Image Completion) aims to synthesize proper contents in missing regions of an image, which can be used in many applications. In simpler words, one can think of Image Inpainting as an image completion technique which aims to present images which look as similar as possible to the original image (the ground truth) and are not easily distinguishable by the human eye. One of the major application of such techniques is in the area of image editing, where one might want to remove some unwanted object from the picture or for the restoration of some cultural heritage image.

Over the years many different approaches have been taken for the process of Image Inpainting. This project aims at using a unified generative network for image inpainting, denoted as dense multi-scale fusion network (DMFN), with self-guided regression loss and geometrical alignment constraint; resulting in highly improved quality of the produced images.

Another goal that this project focuses on is whether Image Inpainting can be used for the data privacy of the visual content (especially images) uploaded by the users on the social media. The primary motivation behind this goal is the rising concerns around the invasive applications of the image data that has piqued the interest of the researchers (and developers) working with the facial recognition tools and technology.