Deep learning-enabled invisibility cloak design

Abstract

Many people have been fascinated with the topic of invisibility since a long time ago and there have been many invisibility cloaks created throughout the years. In this Thesis, a new type of invisibility cloak design is proposed to lessen the effort in creating invisibility cloaks. The proposed invisibility cloak design is entirely made using deep learning models, namely the You only look once (YOLO) detection model and the Guided Language to Image Diffusion for Generation and Editing (GLIDE) generative model. The two models are linked by the creation of a mask, which results in an algorithm that can keep an object out of sight in an image, hence disguising the object.