Removing Unwanted Objects from Image/Frame By Generating Sub-Images Through Generative Adversarial Network

Abstract The amazingly developing and advancing technology in every field also shows itself in image/frame processing. Today, computer vision is an important field of study that can be applied to computers with human vision. Generative networks have been used specifically in studies to specific purpose. After the generating networks were published, they were effectively used to obtain new data for various purposes. Basically, we aimed to remove an unwanted object on the picture or frame. With this aim, thought that like human vision, the foresight of the structures that will replace the object planned to be removed has been considered. Accordingly, if an object is in front of a structure/s the object to desired remove it, this possible data to replace the object must have a similar transitivity with the structure/s behind it. To do this, it is aimed to create new data by using sub-data in object location surrounding the object, with the approach that the appropriate data to replace the object can be filled by learning from the structures around the object. It was tried to replace the object (for pixels containing the object) by generating pixels in object dimensions, using generative networks and useful results were obtained, in this study.