Diminishing unwanted objects based on object detection using deep learning and image inpainting

Abstract

Photography is not done in perfect circumstances. Especially, it is difficult to shoot the only desired object in many circumstances. This paper proposes a method to diminish unwanted objects in photographs. we propose a basic design method of diminished reality using deep learning. Proposed method composed of object detection using deep learning and exemplar-based image inpainting. Previous methods should specify an undesired range before or during operation. Our method has the advantage that it is possible to diminish with the object label to diminish without having to select the area to diminish. The experimental results show naturally diminished scene of the proposed method.